chore: upgrade unregistry 0.3.1 -> 0.4.1 with a fix for pushing big images

* feat: Add `uc service start/stop` commands

* Add nolint:dupl for service start/stop commands

Co-authored-by: Pasha Sviderski <me@psviderski.name>

* Add --signal and --timeout options to uc service stop

* Check whether service stop --timeout flag was used to set options properly

---------

Co-authored-by: Pasha Sviderski <me@psviderski.name>

.github/workflows/go-tests.yml

@@ -28,7 +28,7 @@ jobs:
       - name: Set up Go
         uses: actions/setup-go@d35c59abb061a4a6fb18e82ac0862c26744d6ab5 # v5.5.0
         with:
-          go-version: "1.23.2"
+          go-version: "1.25.1"
 
       - name: Install dependencies
         run: |
@@ -40,7 +40,7 @@ jobs:
         run: |
           make ucind-image
           make test
-    timeout-minutes: 10
+    timeout-minutes: 15
 
   check-protobuf:
     runs-on: ${{ matrix.os }}
@@ -57,7 +57,7 @@ jobs:
       - name: Install Mise
         uses: jdx/mise-action@13abe502c30c1559a5c37dff303831bab82c9402 # v2.2.3
         with:
-          version: "2025.6.5"
+          version: "2025.9.6"
         env:
           GITHUB_TOKEN: ${{ github.token }}
 

.github/workflows/lint.yml

@@ -23,12 +23,12 @@ jobs:
       - name: Set up Go
         uses: actions/setup-go@d35c59abb061a4a6fb18e82ac0862c26744d6ab5 # v5.5.0
         with:
-          go-version: "1.23.2"
+          go-version: "1.25.1"
 
       - name: golangci-lint
         uses: golangci/golangci-lint-action@4afd733a84b1f43292c63897423277bb7f4313a9 # v8.0.0
         with:
-          version: v2.2.2
+          version: v2.4.0
 
       - name: Format code
         run: |

.goreleaser.yaml

@@ -16,7 +16,8 @@ builds:
     main: ./cmd/uncloud
     binary: uncloud
     env:
-      - CGO_ENABLED=0
+      # Use CGO to be able to build fsevent required by compose on macOS.
+      - CGO_ENABLED={{ if eq .Os "darwin" }}1{{ else }}0{{ end }}
     goos:
       - darwin
       - linux

.mise.lock

@@ -1,47 +1,97 @@
-[tools."aqua:vektra/mockery"]
+[[tools."aqua:vektra/mockery"]]
 version = "3.5.3"
 backend = "aqua:vektra/mockery"
 
-[tools."aqua:vektra/mockery".checksums]
-"mockery_3.5.3_Darwin_arm64.tar.gz" = "sha256:a3a94b14c7414e148f2252199ffc4a0108d311358f3d336cbe05bb73cb203704"
-"mockery_3.5.3_Linux_x86_64.tar.gz" = "sha256:ebce416b0175338525246c376885a1579ca6cd4d4015140ba0c70e6b5339a39c"
+[tools."aqua:vektra/mockery".platforms.linux-x64]
+checksum = "sha256:ebce416b0175338525246c376885a1579ca6cd4d4015140ba0c70e6b5339a39c"
+size = 5498859
+url = "https://github.com/vektra/mockery/releases/download/v3.5.3/mockery_3.5.3_Linux_x86_64.tar.gz"
 
-[tools.go]
-version = "1.23.10"
+[tools."aqua:vektra/mockery".platforms.macos-arm64]
+checksum = "sha256:a3a94b14c7414e148f2252199ffc4a0108d311358f3d336cbe05bb73cb203704"
+size = 5238877
+url = "https://github.com/vektra/mockery/releases/download/v3.5.3/mockery_3.5.3_Darwin_arm64.tar.gz"
+
+[[tools.go]]
+version = "1.25.1"
 backend = "core:go"
 
-[tools.go.checksums]
-"go1.23.10.darwin-arm64.tar.gz" = "sha256:25c64bfa8a8fd8e7f62fb54afa4354af8409a4bb2358c2699a1003b733e6fce5"
-"go1.23.10.linux-amd64.tar.gz" = "sha256:535f9f81802499f2a7dbfa70abb8fda3793725fcc29460f719815f6e10b5fd60"
+[tools.go.platforms.linux-x64]
+checksum = "sha256:7716a0d940a0f6ae8e1f3b3f4f36299dc53e31b16840dbd171254312c41ca12e"
+size = 59660846
+url = "https://dl.google.com/go/go1.25.1.linux-amd64.tar.gz"
 
-[tools.golangci-lint]
-version = "2.2.2"
+[tools.go.platforms.macos-arm64]
+checksum = "sha256:68deebb214f39d542e518ebb0598a406ab1b5a22bba8ec9ade9f55fb4dd94a6c"
+size = 57906702
+url = "https://dl.google.com/go/go1.25.1.darwin-arm64.tar.gz"
+
+[[tools.golangci-lint]]
+version = "2.4.0"
 backend = "aqua:golangci/golangci-lint"
 
-[tools.golangci-lint.checksums]
-"golangci-lint-2.2.2-darwin-arm64.tar.gz" = "sha256:d84d94d042c0d495fd1746f3d18948a75de163b17a14e8de3ef840928dd2df74"
-"golangci-lint-2.2.2-linux-amd64.tar.gz" = "sha256:c27fbde948a87d326feacd21df2f61a9c54dbd2e3bfa185c0a1cd6917a6f964f"
+[tools.golangci-lint.platforms.linux-x64]
+checksum = "sha256:fae792524c04424c0ac369f5b8076f04b45cf29fc945a370e55d369a8dc11840"
+size = 14126779
+url = "https://github.com/golangci/golangci-lint/releases/download/v2.4.0/golangci-lint-2.4.0-linux-amd64.tar.gz"
 
-[tools.protoc]
+[tools.golangci-lint.platforms.macos-arm64]
+checksum = "sha256:cd4dd53fa09b6646baff5fd22b8c64d91db02c21c7496df27992d75d34feec59"
+size = 13542526
+url = "https://github.com/golangci/golangci-lint/releases/download/v2.4.0/golangci-lint-2.4.0-darwin-arm64.tar.gz"
+
+[[tools.node]]
+version = "22.20.0"
+backend = "core:node"
+
+[tools.node.platforms.linux-x64]
+checksum = "sha256:eeaccb0378b79406f2208e8b37a62479c70595e20be6b659125eb77dd1ab2a29"
+size = 56645685
+url = "https://nodejs.org/dist/v22.20.0/node-v22.20.0-linux-x64.tar.gz"
+
+[tools.node.platforms.macos-arm64]
+checksum = "sha256:cc04a76a09f79290194c0646f48fec40354d88969bec467789a5d55dd097f949"
+size = 49838299
+url = "https://nodejs.org/dist/v22.20.0/node-v22.20.0-darwin-arm64.tar.gz"
+
+[[tools.protoc]]
 version = "27.3"
 backend = "aqua:protocolbuffers/protobuf/protoc"
 
-[tools.protoc.checksums]
-"protoc-27.3-linux-x86_64.zip" = "sha256:6dab2adab83f915126cab53540d48957c40e9e9023969c3e84d44bfb936c7741"
-"protoc-27.3-osx-aarch_64.zip" = "sha256:b22116bd97cdbd7ea25346abe635a9df268515fe5ef5afa93cd9a68fc2513f84"
+[tools.protoc.platforms.linux-x64]
+checksum = "blake3:8c2938accf7061164ea3968529a47ced1db4824ca553857f6903c68ae4347739"
+size = 3235873
+url = "https://github.com/protocolbuffers/protobuf/releases/download/v27.3/protoc-27.3-linux-x86_64.zip"
 
-[tools.protoc-gen-go]
+[tools.protoc.platforms.macos-arm64]
+checksum = "blake3:650c2494c003e9fddd6f4164a0a36e31da51ae7e1e5208da8cb560a4fe886fcd"
+size = 2372763
+url = "https://github.com/protocolbuffers/protobuf/releases/download/v27.3/protoc-27.3-osx-aarch_64.zip"
+
+[[tools.protoc-gen-go]]
 version = "1.34.2"
 backend = "aqua:protocolbuffers/protobuf-go/protoc-gen-go"
 
-[tools.protoc-gen-go.checksums]
-"protoc-gen-go.v1.34.2.darwin.arm64.tar.gz" = "sha256:17aca7f948dbb624049030cf841e35895cf34183ba006e721247fdeb95ff2780"
-"protoc-gen-go.v1.34.2.linux.amd64.tar.gz" = "sha256:b87bc134dee55576a842141bf0ed27761c635d746780fce5dee038c6dd16554f"
+[tools.protoc-gen-go.platforms.linux-x64]
+checksum = "blake3:f765e393e0f0b25070c13848d88f5e6303e66090132afa1b753e981507a4174e"
+size = 1848331
+url = "https://github.com/protocolbuffers/protobuf-go/releases/download/v1.34.2/protoc-gen-go.v1.34.2.linux.amd64.tar.gz"
 
-[tools.protoc-gen-go-grpc]
+[tools.protoc-gen-go.platforms.macos-arm64]
+checksum = "blake3:c237256a7f718ac4ccdc037ee30025a0b7c3f79300d7831ec59d76d782443fd5"
+size = 1784028
+url = "https://github.com/protocolbuffers/protobuf-go/releases/download/v1.34.2/protoc-gen-go.v1.34.2.darwin.arm64.tar.gz"
+
+[[tools.protoc-gen-go-grpc]]
 version = "1.5.1"
 backend = "aqua:grpc/grpc-go/protoc-gen-go-grpc"
 
-[tools.protoc-gen-go-grpc.checksums]
-"protoc-gen-go-grpc.v1.5.1.darwin.arm64.tar.gz" = "sha256:d6083feb51dcfe59f26793e99ef01ef5eac68b64294ec2546711f614ac5878f3"
-"protoc-gen-go-grpc.v1.5.1.linux.amd64.tar.gz" = "sha256:a6cac4ea731e54aea304ad44d704a69d1cdc82997084b33637e21a89dc9229d6"
+[tools.protoc-gen-go-grpc.platforms.linux-x64]
+checksum = "blake3:197739132f0408f017d0c83773ac1ace85c0572473fdd658ec6e4ccbf20e0552"
+size = 4859466
+url = "https://github.com/grpc/grpc-go/releases/download/cmd/protoc-gen-go-grpc/v1.5.1/protoc-gen-go-grpc.v1.5.1.linux.amd64.tar.gz"
+
+[tools.protoc-gen-go-grpc.platforms.macos-arm64]
+checksum = "blake3:3c668207565d23ea2cf006d699207e3a804ebb929a534bd38a3b4ff4e941e00c"
+size = 4565944
+url = "https://github.com/grpc/grpc-go/releases/download/cmd/protoc-gen-go-grpc/v1.5.1/protoc-gen-go-grpc.v1.5.1.darwin.arm64.tar.gz"

.mise.toml

@@ -3,8 +3,9 @@ experimental = true
 
 [tools]
 "aqua:vektra/mockery" = "3.5.3"
-go = "1.23"
-golangci-lint = "2.2.2"
+go = "1.25.1"
+golangci-lint = "2.4.0"
+node = "22.20.0"
 protoc = "27.3"
 protoc-gen-go = "1.34.2"
 protoc-gen-go-grpc = "1.5.1"

AI.md

@@ -28,7 +28,7 @@ This document provides comprehensive information about the Uncloud project for A
 - **Docker Compose compatibility**: Uses familiar Docker Compose format
 - **Zero-downtime deployments**: Rolling updates without service interruption
 - **Automatic HTTPS**: Caddy reverse proxy with Let's Encrypt integration
-- **Managed DNS**: Free `*.cluster.uncloud.run` subdomains via Uncloud DNS service
+- **Managed DNS**: Free `*.xxxxxx.uncld.dev` subdomains via Uncloud DNS service
 - **Cross-machine scaling**: Run containers across multiple machines
 
 ### 🔧 Developer Experience
@@ -88,6 +88,10 @@ This document provides comprehensive information about the Uncloud project for A
 - **`scripts/`**: Installation and utility scripts
 - **`test/`**: Test suites and test infrastructure
 - **`website/`**: Documentation website (Docusaurus)
+
+  - `landing/`: Landing page
+  - `docs/`: User documentation
+
 - **`misc/`**: Design documents and guides
 
 ## Key Technologies
@@ -184,9 +188,11 @@ uc context use <name>           # Switch context
 
 ### Testing
 
-- Unit tests alongside source files (`*_test.go`)
-- Integration tests in `test/e2e/`
-- Test fixtures in `test/fixtures/`
+- Test files and locations
+  - Unit tests alongside source files (`*_test.go`)
+  - Integration tests in `test/e2e/`
+  - Test fixtures in `test/fixtures/`
+- Use table driven tests whenever possible
 
 ### Dependencies
 
@@ -233,7 +239,7 @@ uc context use <name>           # Switch context
 - `internal/machine/machine.go`: Core machine management
 - `pkg/api/`: Public API definitions
 - `misc/design.md`: Architecture and design philosophy
-- `README.md`: User-facing documentation
+- `README.md`: Repository README
 
 ### Configuration Files
 
@@ -242,3 +248,11 @@ uc context use <name>           # Switch context
 - `Dockerfile`: Container build instructions forUncloud-in-Docker (used for testing)
 
 This document should help AI assistants understand the project structure, make informed suggestions, and contribute effectively to the Uncloud codebase.
+
+## Documentation
+
+Instructions when generating documentation:
+
+- Use conversational language – write as if you were speaking to a friend.
+- Keep sentences simple, optimize for clarity and understanding.
+- Place the subject before the action whenever possible. Example: prefer “The function loads data” over “Data is loaded by the function.”

Dockerfile

@@ -1,6 +1,6 @@
 ARG ALPINE_VERSION=3.20.3
 
-FROM golang:1.23.2-alpine AS uncloudd
+FROM golang:1.25.1-alpine AS uncloudd
 
 ARG TARGETOS
 ARG TARGETARCH
@@ -13,7 +13,7 @@ COPY go.mod go.sum ./
 RUN go mod download && go mod verify
 
 COPY . .
-RUN GOOS=${TARGETOS} GOARCH=${TARGETARCH} go build -o uncloudd cmd/uncloudd/main.go
+RUN GOOS=${TARGETOS} GOARCH=${TARGETARCH} go build -o uncloudd ./cmd/uncloudd
 
 
 FROM alpine:${ALPINE_VERSION} AS corrosion-download

Makefile

@@ -38,7 +38,7 @@ demo-reset:
 
 .PHONY: ucind-cluster
 ucind-cluster:
-	go run ./cmd/ucind cluster rm && go run ./cmd/ucind cluster create -m 3
+	go run ./cmd/ucind cluster rm && go run ./cmd/ucind cluster create -m $(if $(MACHINES_COUNT),$(MACHINES_COUNT),3)
 
 .PHONY: proto
 proto:

README.md

@@ -1,7 +1,7 @@
 <div align="center">
-  <img src="./website/landing/images/logo.svg" height="100" width="100" alt="Uncloud logo"/>
-  <h1>Uncloud</h1>
-  <p><strong>Docker simplicity. Multi-machine power.</strong></p>
+  <img src="./website/landing/images/logo-title.svg#gh-light-mode-only" alt="Uncloud logo"/>
+  <img src="./website/landing/images/logo-title-dark.svg#gh-dark-mode-only" alt="Uncloud logo"/>
+  <p><strong>▸ Docker simplicity. Multi-machine power ◂</strong></p>
 
   <p>
     <a href="https://uncloud.run/docs"><img src="https://img.shields.io/badge/Docs-blue.svg?style=for-the-badge&logo=gitbook&logoColor=white" alt="Documentation"></a>
@@ -25,12 +25,14 @@ complexity of Kubernetes.
 
 ## ✨ Features
 
-* **Deploy anywhere**: Combine cloud VMs, dedicated servers, and bare metal into a unified computing environment —
+* **Deploy anywhere**: Combine cloud VMs, dedicated servers, and bare metal into a unified computing environment,
   regardless of location or provider.
 * **Docker Compose**: Familiar [Docker Compose](https://compose-spec.io/) format for defining services and volumes. No
   need to learn a new bespoke DSL.
 * **Zero-downtime deployments**: Rolling updates without service interruption. Automatic rollback on failure is coming
   soon.
+* **[Unregistry](https://github.com/psviderski/unregistry) integration**: Build and push your Docker images directly to
+  your machines without an external registry. It will transfer only the missing layers, making it fast and efficient.
 * **Service discovery**: Built-in DNS server resolves service names to container IPs.
 * **Persistent storage**: Run stateful services with Docker volumes managed across machines.
 * **Zero-config private network**: Automatic WireGuard mesh with peer discovery and NAT traversal. Containers get unique
@@ -38,18 +40,13 @@ complexity of Kubernetes.
 * **No control plane**: Fully decentralised design eliminates single points of failure and reduces operational overhead.
 * **Imperative over declarative**: Favoring imperative operations over state reconciliation simplifies both the mental
   model and troubleshooting.
-* **Managed DNS**: Automatic DNS records `*.<id>.cluster.uncloud.run` for services with public access via managed
+* **Managed DNS**: Automatic DNS records `*.xxxxxx.uncld.dev` for services with public access via managed
   [Uncloud DNS](https://github.com/psviderski/uncloud-dns) service.
 * **Automatic HTTPS**: Built-in Caddy reverse proxy handles TLS certificate provisioning and renewal using Let's
   Encrypt.
 * **Docker-like CLI**: Familiar commands for managing both infrastructure and applications.
 * **Remote management**: Control your entire infrastructure through SSH access to any single machine in the cluster.
 
-### 🚀 Coming soon
-
-* **[Unregistry](https://github.com/psviderski/unregistry) integration**: Push your Docker images directly to your
-  machines without an external registry. It will transfer only the missing layers, making it fast and efficient.
-
 ## 🎬 Quick demo
 
 The screenshot below demonstrates how I use Uncloud to deploy https://uncloud.run website to 2 remote machines from
@@ -67,6 +64,12 @@ on-premises in just a couple minutes.
 <img src="https://embed-ssl.wistia.com/deliveries/3cf7014a48b93afc556444bed3e39a8c.jpg?image_crop_resized=900x526&image_play_button_rounded=true&image_play_button_size=2x&image_play_button_color=18181Be0" alt="Uncloud demo" width="450" height="263" />
 </a>
 
+<br>
+<br>
+
+> **📚 Want more examples?** Check out the [**uncloud-recipes**](https://github.com/psviderski/uncloud-recipes)
+> repository for community recipes and templates for deploying popular services on Uncloud.
+
 ## 💫 Why Uncloud?
 
 Modern cloud platforms like Heroku and Render offer amazing developer experiences but at a premium price. Traditional
@@ -334,6 +337,10 @@ These companies and projects are helping Uncloud with their generous sponsorship
 
 ## ❤️ Contributors
 
+<a href="https://trendshift.io/repositories/14069">
+    <img alt="Trendshift" src="https://trendshift.io/api/badge/repositories/14069" />
+</a>
+
 Thank you [@cedws](https://github.com/cedws) for being the first contributor to Uncloud! 🎉
 
 <a href="https://github.com/psviderski/uncloud/graphs/contributors">

cmd/uncloud/build.go

@@ -10,68 +10,102 @@ import (
 	"github.com/spf13/cobra"
 )
 
-// NewBuildCommand creates a new command to build services from a Compose file.
+type buildOptions struct {
+	cli.BuildServicesOptions
+	files    []string
+	profiles []string
+}
+
+// NewBuildCommand creates a new command to build images for services from a Compose file.
 func NewBuildCommand() *cobra.Command {
-	opts := cli.BuildOptions{}
+	opts := buildOptions{}
 	cmd := &cobra.Command{
 		Use:   "build [FLAGS] [SERVICE...]",
 		Short: "Build services from a Compose file.",
+		Long: `Build images for services from a Compose file using local Docker.
+
+By default, built images remain on the local Docker host. Use --push to upload them
+to cluster machines or --push-registry to upload them to external registries.`,
+		Example: `  # Build all services that have a build section in compose.yaml.
+  uc build
+
+  # Build specific services that have a build section.
+  uc build web api
+
+  # Build services and push images to all cluster machines or service x-machines if specified.
+  uc build --push
+
+  # Build services and push images to specific machines.
+  uc build --push -m machine1,machine2
+
+  # Build services and push images to external registries (e.g., Docker Hub).
+  uc build --push-registry
+
+  # Build services with build arguments, pull newer base images before building, and don't use cache.
+  uc build --build-arg NODE_VERSION=24 --build-arg ENV=production --no-cache --pull`,
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-
-			if len(args) > 0 {
 			opts.Services = args
-			}
 
 			return runBuild(cmd.Context(), uncli, opts)
 		},
+		GroupID: "service",
 	}
 
-	cmd.Flags().StringSliceVarP(&opts.Files, "file", "f", nil,
-		"One or more Compose files to build (default compose.yaml)")
-	cmd.Flags().StringSliceVarP(&opts.Profiles, "profile", "p", nil,
+	cmd.Flags().StringArrayVar(&opts.BuildArgs, "build-arg", nil,
+		"Set a build-time variable for services. Used in Dockerfiles that declare the variable with ARG.\n"+
+			"Can be specified multiple times. Format: --build-arg VAR=VALUE")
+	cmd.Flags().BoolVar(&opts.Check, "check", false,
+		"Check the build configuration for services without building them.")
+	cmd.Flags().BoolVar(&opts.Deps, "deps", false,
+		"Also build services declared as dependencies of the selected services.")
+	cmd.Flags().StringSliceVarP(&opts.files, "file", "f", nil,
+		"One or more Compose files to build. (default compose.yaml)")
+	cmd.Flags().StringSliceVarP(&opts.Machines, "machine", "m", nil,
+		"Machine names or IDs to push the built images to (requires --push).\n"+
+			"Can be specified multiple times or as a comma-separated list. (default is all machines or x-machines)")
+	cmd.Flags().BoolVar(&opts.NoCache, "no-cache", false,
+		"Do not use cache when building images.")
+	cmd.Flags().StringSliceVarP(&opts.profiles, "profile", "p", nil,
 		"One or more Compose profiles to enable.")
-	cmd.Flags().BoolVarP(&opts.Push, "push", "P", false,
-		"Push built images to the registry after building. (default false)")
-	cmd.Flags().BoolVarP(&opts.NoCache, "no-cache", "n", false,
-		"Do not use cache when building images. (default false)")
+	cmd.Flags().BoolVar(&opts.Pull, "pull", false,
+		"Always attempt to pull newer versions of base images before building.")
+	cmd.Flags().BoolVar(&opts.PushCluster, "push", false,
+		"Upload the built images to cluster machines after building.\n"+
+			"Use --machine to specify which machines. (default is all machines)")
+	cmd.Flags().BoolVar(&opts.PushRegistry, "push-registry", false,
+		"Upload the built images to external registries (e.g., Docker Hub) after building.")
 
 	return cmd
 }
 
-// TODO: deduplicate with a similar functino for deploy options
-// projectOpts returns the project options for the Compose file(s).
-func projectOptsFromBuildOpts(opts cli.BuildOptions) []composecli.ProjectOptionsFn {
-	projectOpts := []composecli.ProjectOptionsFn{}
-
-	if len(opts.Profiles) > 0 {
-		projectOpts = append(projectOpts, composecli.WithDefaultProfiles(opts.Profiles...))
+// runBuild parses the Compose file(s) and builds the images for selected services.
+func runBuild(ctx context.Context, uncli *cli.CLI, opts buildOptions) error {
+	// Validate push flags.
+	if opts.PushCluster && opts.PushRegistry {
+		return fmt.Errorf("cannot specify both --push and --push-registry: choose one push target")
 	}
 
-	return projectOpts
-}
+	machines := cli.ExpandCommaSeparatedValues(opts.Machines)
+	// Special handling for an explicit "all" keyword to push to all machines.
+	if len(machines) == 1 && machines[0] == "all" {
+		machines = nil
+	}
+	opts.Machines = machines
 
-// runBuild parses the Compose file(s), builds the services, and pushes them if requested.
-func runBuild(ctx context.Context, uncli *cli.CLI, opts cli.BuildOptions) error {
-	projectOpts := projectOptsFromBuildOpts(opts)
-	project, err := compose.LoadProject(ctx, opts.Files, projectOpts...)
+	project, err := compose.LoadProject(ctx, opts.files, composecli.WithDefaultProfiles(opts.profiles...))
 	if err != nil {
 		return fmt.Errorf("load compose file(s): %w", err)
 	}
 
-	if len(opts.Services) > 0 {
-		project, err = project.WithSelectedServices(opts.Services)
+	servicesToBuild, err := cli.ServicesThatNeedBuild(project, opts.Services, opts.Deps)
 	if err != nil {
-			return fmt.Errorf("select services: %w", err)
+		return fmt.Errorf("determine services to build: %w", err)
 	}
-	}
-
-	servicesToBuild := cli.GetServicesThatNeedBuild(project)
-
 	if len(servicesToBuild) == 0 {
 		fmt.Println("No services to build.")
 		return nil
 	}
 
-	return cli.BuildServices(ctx, servicesToBuild, opts)
+	return uncli.BuildServices(ctx, project, opts.BuildServicesOptions)
 }

cmd/uncloud/caddy/config.go

@@ -7,14 +7,12 @@ import (
 
 	"github.com/alecthomas/chroma/v2/quick"
 	"github.com/psviderski/uncloud/internal/cli"
-	"github.com/psviderski/uncloud/pkg/api"
 	"github.com/spf13/cobra"
 )
 
 type configOptions struct {
 	machine string
 	noColor bool
-	context string
 }
 
 func NewConfigCommand() *cobra.Command {
@@ -34,16 +32,12 @@ func NewConfigCommand() *cobra.Command {
 		"Name or ID of the machine to get the configuration from. (default is connected machine)")
 	cmd.Flags().BoolVar(&opts.noColor, "no-color", false,
 		"Disable syntax highlighting for the output.")
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 
 	return cmd
 }
 
 func runConfig(ctx context.Context, uncli *cli.CLI, opts configOptions) error {
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
@@ -51,7 +45,7 @@ func runConfig(ctx context.Context, uncli *cli.CLI, opts configOptions) error {
 
 	if opts.machine != "" {
 		// If a specific machine is requested, use it to get the Caddy configuration.
-		ctx, _, err = api.ProxyMachinesContext(ctx, clusterClient, []string{opts.machine})
+		ctx, _, err = clusterClient.ProxyMachinesContext(ctx, []string{opts.machine})
 		if err != nil {
 			return err
 		}

cmd/uncloud/caddy/deploy.go

@@ -22,7 +22,6 @@ type deployOptions struct {
 	caddyfile string
 	image     string
 	machines  []string
-	context   string
 }
 
 func NewDeployCommand() *cobra.Command {
@@ -44,12 +43,8 @@ func NewDeployCommand() *cobra.Command {
 	cmd.Flags().StringVar(&opts.image, "image", "",
 		"Caddy Docker image to deploy. (default caddy:LATEST_VERSION)")
 	cmd.Flags().StringSliceVarP(&opts.machines, "machine", "m", nil,
-		"Machine names to deploy to. Can be specified multiple times or as a comma-separated "+
-			"list of machine names. (default is all machines)")
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context to deploy to. (default is the current context)",
-	)
+		"Machine names or IDs to deploy to. Can be specified multiple times or as a comma-separated "+
+			"list. (default is all machines)")
 
 	return cmd
 }
@@ -64,7 +59,7 @@ func runDeploy(ctx context.Context, uncli *cli.CLI, opts deployOptions) error {
 		caddyfile = strings.TrimSpace(string(data))
 	}
 
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/context/connection.go

@@ -0,0 +1,78 @@
+package context
+
+import (
+	"fmt"
+
+	"github.com/charmbracelet/huh"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/internal/cli/config"
+	"github.com/spf13/cobra"
+)
+
+func NewConnectionCommand() *cobra.Command {
+	cmd := &cobra.Command{
+		Use:   "connection",
+		Short: "Choose a new default connection for the current context.",
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			return selectConnection(uncli)
+		},
+	}
+	return cmd
+}
+
+func selectConnection(uncli *cli.CLI) error {
+	if uncli.Config == nil {
+		return fmt.Errorf("connection management is not available: Uncloud configuration file is not being used")
+	}
+	if len(uncli.Config.Contexts) == 0 {
+		return fmt.Errorf("no contexts found in Uncloud config (%s)", uncli.Config.Path())
+	}
+
+	currentCtxName := uncli.Config.CurrentContext
+	currentCtx, ok := uncli.Config.Contexts[currentCtxName]
+	if !ok {
+		return fmt.Errorf("current context '%s' not found", currentCtxName)
+	}
+
+	if len(currentCtx.Connections) == 0 {
+		return fmt.Errorf("no connections found in context '%s'", currentCtxName)
+	}
+
+	var selectedConnection int
+	form := huh.NewForm(
+		huh.NewGroup(
+			huh.NewSelect[int]().
+				Title("Select a default connection").
+				Options(buildConnectionOptions(currentCtx.Connections)...).
+				Value(&selectedConnection),
+		),
+	)
+	if err := form.Run(); err != nil {
+		return fmt.Errorf("select connection: %w", err)
+	}
+
+	currentCtx.SetDefaultConnection(selectedConnection)
+	selected := currentCtx.Connections[0]
+
+	if err := uncli.Config.Save(); err != nil {
+		return fmt.Errorf("save config: %w", err)
+	}
+
+	fmt.Printf("Default connection for context '%s' is now '%s'.\n", currentCtxName, selected)
+	return nil
+}
+
+func buildConnectionOptions(connections []config.MachineConnection) []huh.Option[int] {
+	options := make([]huh.Option[int], len(connections))
+	for i, conn := range connections {
+		key := conn.String()
+		opt := huh.NewOption(key, i)
+		if i == 0 {
+			opt.Key += " (default)"
+			opt = opt.Selected(true)
+		}
+		options[i] = opt
+	}
+	return options
+}

cmd/uncloud/context/ls.go

@@ -12,8 +12,6 @@ import (
 )
 
 func NewListCommand() *cobra.Command {
-	var clusterContext string
-
 	cmd := &cobra.Command{
 		Use:     "ls",
 		Aliases: []string{"list"},
@@ -24,11 +22,6 @@ func NewListCommand() *cobra.Command {
 		},
 	}
 
-	cmd.Flags().StringVarP(
-		&clusterContext, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
-
 	return cmd
 }
 
@@ -51,7 +44,7 @@ func list(uncli *cli.CLI) error {
 	for _, name := range contextNames {
 		current := ""
 		if name == currentContext {
-			current = "*"
+			current = "✓"
 		}
 		connCount := len(uncli.Config.Contexts[name].Connections)
 		fmt.Fprintf(tw, "%s\t%s\t%d\n", name, current, connCount)

cmd/uncloud/context/root.go

@@ -19,6 +19,7 @@ func NewRootCommand() *cobra.Command {
 	cmd.AddCommand(
 		NewListCommand(),
 		NewUseCommand(),
+		NewConnectionCommand(),
 	)
 
 	return cmd

cmd/uncloud/deploy.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"strings"
 
+	"github.com/charmbracelet/lipgloss"
 	composecli "github.com/compose-spec/compose-go/v2/cli"
 	"github.com/docker/compose/v2/pkg/progress"
 	"github.com/psviderski/uncloud/internal/cli"
@@ -17,13 +18,14 @@ import (
 )
 
 type deployOptions struct {
+	cli.BuildServicesOptions
+
 	files    []string
 	profiles []string
 	services []string
 	noBuild  bool
 	recreate bool
-
-	context string
+	yes      bool
 }
 
 // NewDeployCommand creates a new command to deploy services from a Compose file.
@@ -33,26 +35,34 @@ func NewDeployCommand() *cobra.Command {
 		Use:   "deploy [FLAGS] [SERVICE...]",
 		Short: "Deploy services from a Compose file.",
 		RunE: func(cmd *cobra.Command, args []string) error {
-			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			cli.BindEnvToFlag(cmd, "yes", "UNCLOUD_AUTO_CONFIRM")
 
-			if len(args) > 0 {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
 			opts.services = args
-			}
 
 			return runDeploy(cmd.Context(), uncli, opts)
 		},
+		GroupID: "service",
 	}
 
+	cmd.Flags().StringArrayVar(&opts.BuildServicesOptions.BuildArgs, "build-arg", nil,
+		"Set a build-time variable for services. Used in Dockerfiles that declare the variable with ARG.\n"+
+			"Can be specified multiple times. Format: --build-arg VAR=VALUE")
+	cmd.Flags().BoolVar(&opts.BuildServicesOptions.Pull, "build-pull", false,
+		"Always attempt to pull newer versions of base images before building service images.")
 	cmd.Flags().StringSliceVarP(&opts.files, "file", "f", nil,
 		"One or more Compose files to deploy services from. (default compose.yaml)")
+	cmd.Flags().BoolVar(&opts.noBuild, "no-build", false,
+		"Do not build new images before deploying services.")
+	cmd.Flags().BoolVar(&opts.BuildServicesOptions.NoCache, "no-cache", false,
+		"Do not use cache when building images.")
 	cmd.Flags().StringSliceVarP(&opts.profiles, "profile", "p", nil,
 		"One or more Compose profiles to enable.")
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context to deploy to (default is the current context)")
-	cmd.Flags().BoolVarP(&opts.noBuild, "no-build", "n", false,
-		"Do not build images before deploying services. (default false)")
 	cmd.Flags().BoolVar(&opts.recreate, "recreate", false,
 		"Recreate containers even if their configuration and image haven't changed.")
+	cmd.Flags().BoolVarP(&opts.yes, "yes", "y", false,
+		"Auto-confirm deployment plan. Should be explicitly set when running non-interactively,\n"+
+			"e.g., in CI/CD pipelines. [$UNCLOUD_AUTO_CONFIRM]")
 
 	// TODO: Consider adding a filter flag to specify which machines to deploy to but keep the rest running.
 	//  Could be useful to test a new version on a subset of machines before rolling out to all.
@@ -60,22 +70,9 @@ func NewDeployCommand() *cobra.Command {
 	return cmd
 }
 
-// projectOpts returns the project options for the Compose file(s).
-func projectOpts(opts deployOptions) []composecli.ProjectOptionsFn {
-	projectOpts := []composecli.ProjectOptionsFn{}
-
-	if len(opts.profiles) > 0 {
-		projectOpts = append(projectOpts, composecli.WithDefaultProfiles(opts.profiles...))
-	}
-
-	return projectOpts
-}
-
 // runDeploy parses the Compose file(s) and deploys the services.
 func runDeploy(ctx context.Context, uncli *cli.CLI, opts deployOptions) error {
-	projectOpts := projectOpts(opts)
-
-	project, err := compose.LoadProject(ctx, opts.files, projectOpts...)
+	project, err := compose.LoadProject(ctx, opts.files, composecli.WithDefaultProfiles(opts.profiles...))
 	if err != nil {
 		return fmt.Errorf("load compose file(s): %w", err)
 	}
@@ -88,29 +85,69 @@ func runDeploy(ctx context.Context, uncli *cli.CLI, opts deployOptions) error {
 		}
 	}
 
-	servicesToBuild := cli.GetServicesThatNeedBuild(project)
+	servicesToBuild, err := cli.ServicesThatNeedBuild(project, opts.services, false)
+	if err != nil {
+		return fmt.Errorf("determine services to build: %w", err)
+	}
 
 	if len(servicesToBuild) > 0 {
 		if opts.noBuild {
 			fmt.Println("Not building services as requested.")
 		} else {
-			buildOpts := cli.BuildOptions{
-				Push:    true,
-				NoCache: false,
-			}
+			// Build service images without pushing them to cluster yet to not connect to the cluster twice.
+			opts.BuildServicesOptions.Deps = true // build dependencies as deploy includes them by default
+			opts.BuildServicesOptions.Services = opts.services
 
-			if err := cli.BuildServices(ctx, servicesToBuild, buildOpts); err != nil {
+			if err = uncli.BuildServices(ctx, project, opts.BuildServicesOptions); err != nil {
 				return fmt.Errorf("build services: %w", err)
 			}
 		}
+		fmt.Println()
 	}
 
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
 	defer clusterClient.Close()
 
+	if len(servicesToBuild) > 0 && !opts.noBuild {
+		// Push built service images to cluster machines one at a time.
+		var errs []error
+		for _, s := range servicesToBuild {
+			if s.Image == "" {
+				// Skip services without an image name (shouldn't happen for services with build config).
+				continue
+			}
+
+			// Push to the specified x-machines or to *all* cluster machines if not specified.
+			var pushOpts client.PushImageOptions
+			if machines, ok := s.Extensions[compose.MachinesExtensionKey].(compose.MachinesSource); ok {
+				pushOpts.Machines = machines
+			}
+			if len(pushOpts.Machines) == 0 {
+				pushOpts.AllMachines = true
+			}
+
+			boldStyle := lipgloss.NewStyle().Bold(true)
+			err = progress.RunWithTitle(ctx, func(ctx context.Context) error {
+				if err = clusterClient.PushImage(ctx, s.Image, pushOpts); err != nil {
+					return fmt.Errorf("push image '%s' for service '%s': %w", s.Image, s.Name, err)
+				}
+				return nil
+			}, uncli.ProgressOut(), fmt.Sprintf("Pushing image %s to cluster", boldStyle.Render(s.Image)))
+			// Collect errors to try pushing all images.
+			if err != nil {
+				errs = append(errs, err)
+			}
+		}
+
+		if err = errors.Join(errs...); err != nil {
+			return err
+		}
+		fmt.Println()
+	}
+
 	var strategy deploy.Strategy
 	if opts.recreate {
 		strategy = &deploy.RollingStrategy{ForceRecreate: true}
@@ -130,12 +167,19 @@ func runDeploy(ctx context.Context, uncli *cli.CLI, opts deployOptions) error {
 		return nil
 	}
 
-	fmt.Println("Deployment plan:")
+	fmt.Println(lipgloss.NewStyle().Bold(true).Render("Deployment plan"))
 	if err = printPlan(ctx, clusterClient, plan); err != nil {
 		return fmt.Errorf("print deployment plan: %w", err)
 	}
 	fmt.Println()
 
+	// Ask for plan confirmation before proceeding with the deployment unless auto-confirmed with --yes.
+	if !opts.yes {
+		if !cli.IsStdinTerminal() {
+			return errors.New("cannot ask to confirm deployment plan in non-interactive mode, " +
+				"use --yes flag or set UNCLOUD_AUTO_CONFIRM=true to auto-confirm")
+		}
+
 		confirmed, err := cli.Confirm()
 		if err != nil {
 			return fmt.Errorf("confirm deployment: %w", err)
@@ -144,6 +188,7 @@ func runDeploy(ctx context.Context, uncli *cli.CLI, opts deployOptions) error {
 			fmt.Println("Cancelled. No changes were made.")
 			return nil
 		}
+	}
 
 	return progress.RunWithTitle(ctx, func(ctx context.Context) error {
 		if err := plan.Execute(ctx, clusterClient); err != nil {

cmd/uncloud/dns/release.go

@@ -12,32 +12,21 @@ import (
 	"google.golang.org/protobuf/types/known/emptypb"
 )
 
-type releaseOptions struct {
-	context string
-}
-
 func NewReleaseCommand() *cobra.Command {
-	opts := releaseOptions{}
-
 	cmd := &cobra.Command{
 		Use:   "release",
 		Short: "Release the reserved cluster domain.",
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-			return release(cmd.Context(), uncli, opts)
+			return release(cmd.Context(), uncli)
 		},
 	}
 
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
-
 	return cmd
 }
 
-func release(ctx context.Context, uncli *cli.CLI, opts releaseOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+func release(ctx context.Context, uncli *cli.CLI) error {
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/dns/reserve.go

@@ -19,7 +19,6 @@ const DefaultUncloudDNSAPIEndpoint = "https://dns.uncloud.run/v1"
 
 type reserveOptions struct {
 	endpoint string
-	context  string
 }
 
 func NewReserveCommand() *cobra.Command {
@@ -36,16 +35,12 @@ func NewReserveCommand() *cobra.Command {
 
 	cmd.Flags().StringVar(&opts.endpoint, "endpoint", DefaultUncloudDNSAPIEndpoint,
 		"API endpoint for the Uncloud DNS service.")
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 
 	return cmd
 }
 
 func reserve(ctx context.Context, uncli *cli.CLI, opts reserveOptions) error {
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/dns/root.go

@@ -9,7 +9,7 @@ func NewRootCommand() *cobra.Command {
 		Use:   "dns",
 		Short: "Manage cluster domain in Uncloud DNS.",
 		Long: "Manage cluster domain in Uncloud DNS.\n" +
-			"DNS commands allow you to reserve or release a unique '<id>.cluster.uncloud.run' domain for your " +
+			"DNS commands allow you to reserve or release a unique 'xxxxxx.uncld.dev' domain for your " +
 			"cluster. When reserved, Caddy service deployments will automatically update DNS records to route " +
 			"traffic to the services in the cluster.",
 	}

cmd/uncloud/dns/show.go

@@ -10,32 +10,21 @@ import (
 	"github.com/spf13/cobra"
 )
 
-type showOptions struct {
-	context string
-}
-
 func NewShowCommand() *cobra.Command {
-	opts := showOptions{}
-
 	cmd := &cobra.Command{
 		Use:   "show",
 		Short: "Print the cluster domain name.",
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-			return show(cmd.Context(), uncli, opts)
+			return show(cmd.Context(), uncli)
 		},
 	}
 
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
-
 	return cmd
 }
 
-func show(ctx context.Context, uncli *cli.CLI, opts showOptions) error {
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+func show(ctx context.Context, uncli *cli.CLI) error {
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/image/ls.go

@@ -0,0 +1,307 @@
+package image
+
+import (
+	"context"
+	"fmt"
+	"slices"
+	"sort"
+	"strings"
+	"time"
+
+	"github.com/charmbracelet/lipgloss"
+	"github.com/charmbracelet/lipgloss/table"
+	"github.com/containerd/platforms"
+	"github.com/docker/docker/api/types/image"
+	"github.com/docker/go-units"
+	"github.com/muesli/termenv"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/api"
+	"github.com/spf13/cobra"
+)
+
+type listOptions struct {
+	machines   []string
+	nameFilter string
+}
+
+func NewListCommand() *cobra.Command {
+	opts := listOptions{}
+
+	cmd := &cobra.Command{
+		Use:     "ls [REPO:[TAG]]",
+		Aliases: []string{"list"},
+		Short:   "List images on machines in the cluster.",
+		Long:    "List images on machines in the cluster. By default, on all machines. Optionally filter by image name.",
+		Example: `  # List all images on all machines.
+  uc image ls
+
+  # List images on specific machine.
+  uc image ls -m machine1
+
+  # List images on multiple machines.
+  uc image ls -m machine1,machine2
+
+  # List images filtered by name (with any tag) on all machines.
+  uc image ls myapp
+
+  # List images filtered by name pattern on specific machine.
+  uc image ls "myapp:1.*" -m machine1`,
+		Args: cobra.MaximumNArgs(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if len(args) > 0 {
+				opts.nameFilter = args[0]
+			}
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			return list(cmd.Context(), uncli, opts)
+		},
+	}
+
+	cmd.Flags().StringSliceVarP(&opts.machines, "machine", "m", nil,
+		"Filter images by machine name or ID. Can be specified multiple times or as a comma-separated list. "+
+			"(default is include all machines)")
+
+	return cmd
+}
+
+// imageRow represents a single image with its metadata for display.
+type imageRow struct {
+	id           string
+	name         string
+	platforms    string
+	createdHuman string
+	createdUnix  int64
+	size         string
+	inUse        string
+	store        string
+	machine      string
+}
+
+func list(ctx context.Context, uncli *cli.CLI, opts listOptions) error {
+	clusterClient, err := uncli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer clusterClient.Close()
+
+	// Get all machines to create ID to name mapping.
+	allMachines, err := clusterClient.ListMachines(ctx, nil)
+	if err != nil {
+		return fmt.Errorf("list machines: %w", err)
+	}
+
+	machineIDToName := make(map[string]string)
+	for _, machineMember := range allMachines {
+		if machineMember.Machine != nil && machineMember.Machine.Id != "" && machineMember.Machine.Name != "" {
+			machineIDToName[machineMember.Machine.Id] = machineMember.Machine.Name
+		}
+	}
+
+	machines := cli.ExpandCommaSeparatedValues(opts.machines)
+
+	clusterImages, err := clusterClient.ListImages(ctx, api.ImageFilter{
+		Machines: machines,
+		Name:     opts.nameFilter,
+	})
+	if err != nil {
+		return fmt.Errorf("list images: %w", err)
+	}
+
+	// Collect all images from all machines.
+	var rows []imageRow
+
+	for _, machineImages := range clusterImages {
+		// Get machine name for better readability.
+		machineName := machineImages.Metadata.Machine
+		if m := allMachines.FindByNameOrID(machineName); m != nil {
+			machineName = m.Machine.Name
+		}
+
+		store := "docker"
+		if machineImages.ContainerdStore {
+			store = "containerd"
+		}
+
+		// Process each image for this machine.
+		for _, img := range machineImages.Images {
+			// Show the first 12 chars without 'sha256:' as the image ID like Docker does.
+			id := strings.TrimPrefix(img.ID, "sha256:")[:12]
+
+			name := "<none>"
+			if len(img.RepoTags) > 0 && img.RepoTags[0] != "<none>:<none>" {
+				name = img.RepoTags[0]
+			}
+
+			imgPlatforms, _ := imagePlatforms(img)
+			formattedPlatforms := formatPlatforms(imgPlatforms)
+
+			created := ""
+			createdAt := time.Unix(img.Created, 0)
+			if !createdAt.IsZero() {
+				created = units.HumanDuration(time.Now().UTC().Sub(createdAt)) + " ago"
+			}
+
+			size := units.HumanSizeWithPrecision(float64(img.Size), 3)
+
+			// Check if the image is in use by any containers. Only supported by Docker API >=1.51
+			inUse := "-"
+			if img.Containers != -1 { // -1 means the info is not available.
+				if img.Containers > 0 {
+					inUse = lipgloss.NewStyle().Foreground(lipgloss.Color("10")).Render("●")
+				} else {
+					inUse = lipgloss.NewStyle().Foreground(lipgloss.Color("8")).Render("○")
+				}
+			}
+
+			rows = append(rows, imageRow{
+				id:           id,
+				name:         name,
+				platforms:    formattedPlatforms,
+				createdHuman: created,
+				createdUnix:  img.Created,
+				size:         size,
+				inUse:        inUse,
+				store:        store,
+				machine:      machineName,
+			})
+		}
+	}
+
+	if len(rows) == 0 {
+		if opts.nameFilter != "" {
+			fmt.Printf("No images matching '%s' found.\n", opts.nameFilter)
+		} else {
+			fmt.Println("No images found.")
+		}
+		return nil
+	}
+
+	// Sort images by name, then by machine name.
+	sort.Slice(rows, func(i, j int) bool {
+		if rows[i].name != rows[j].name {
+			return rows[i].name < rows[j].name
+		}
+		return rows[i].machine < rows[j].machine
+	})
+
+	// Print the images in a table format.
+	fmt.Println(formatImageTable(rows))
+
+	return nil
+}
+
+// imagePlatforms returns a list of platforms supported by the image and a boolean indicating if it's multi-platform.
+func imagePlatforms(img image.Summary) ([]string, bool) {
+	var formattedPlatforms []string
+	multiPlatform := false
+
+	for _, m := range img.Manifests {
+		if m.Kind != image.ManifestKindImage || !m.Available {
+			continue
+		}
+
+		if m.ID != img.ID {
+			// There is an image manifest that has digest different from the main image digest.
+			// This means the image manifest is an index or a manifest list (multi-platform image).
+			multiPlatform = true
+		}
+		formattedPlatforms = append(formattedPlatforms, platforms.Format(m.ImageData.Platform))
+	}
+
+	slices.Sort(formattedPlatforms)
+
+	return formattedPlatforms, multiPlatform
+}
+
+func formatPlatforms(platforms []string) string {
+	if len(platforms) == 0 {
+		return "-"
+	}
+
+	platformStyle := lipgloss.NewStyle().
+		BorderForeground(lipgloss.Color("152")).
+		Foreground(lipgloss.Color("0")).
+		Background(lipgloss.Color("152"))
+	// Use fancy pill borders only if the output is a terminal with color support.
+	if lipgloss.ColorProfile() != termenv.Ascii {
+		platformStyle = platformStyle.Border(lipgloss.Border{Left: "", Right: ""}, false, true, false, true)
+	}
+
+	styledPlatforms := make([]string, len(platforms))
+	for i, p := range platforms {
+		styledPlatforms[i] = platformStyle.Render(p)
+	}
+
+	return strings.Join(styledPlatforms, " ")
+}
+
+func formatImageTable(rows []imageRow) string {
+	columns := []struct {
+		name string
+		hide bool
+	}{
+		{name: "IMAGE ID"},
+		{name: "NAME"},
+		{name: "PLATFORMS"},
+		{name: "CREATED"},
+		{name: "SIZE"},
+		{name: "IN USE"},
+		{name: "STORE"},
+		{name: "MACHINE"},
+	}
+
+	// Hide the "IN USE" column if none of the images have that info available.
+	inUseInfoAvailable := slices.ContainsFunc(rows, func(r imageRow) bool {
+		return r.inUse != "-"
+	})
+	if !inUseInfoAvailable {
+		// Hide "IN USE" column.
+		columns[5].hide = true
+	}
+
+	t := table.New().
+		// Remove the default border.
+		Border(lipgloss.Border{}).
+		BorderTop(false).
+		BorderBottom(false).
+		BorderLeft(false).
+		BorderRight(false).
+		BorderHeader(false).
+		BorderColumn(false).
+		StyleFunc(func(row, col int) lipgloss.Style {
+			if row == table.HeaderRow {
+				return lipgloss.NewStyle().Bold(true).PaddingRight(3)
+			}
+			// Regular style for data rows with padding.
+			return lipgloss.NewStyle().PaddingRight(3)
+		})
+
+	var headers []string
+	for _, col := range columns {
+		if !col.hide {
+			headers = append(headers, col.name)
+		}
+	}
+	t.Headers(headers...)
+
+	for _, row := range rows {
+		values := []string{
+			row.id,
+			row.name,
+			row.platforms,
+			row.createdHuman,
+			row.size,
+			row.inUse,
+			row.store,
+			row.machine,
+		}
+		var filteredValues []string
+		for i, v := range values {
+			if !columns[i].hide {
+				filteredValues = append(filteredValues, v)
+			}
+		}
+		t.Row(filteredValues...)
+	}
+
+	return t.String()
+}

cmd/uncloud/image/push.go

@@ -0,0 +1,94 @@
+package image
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/containerd/platforms"
+	"github.com/docker/compose/v2/pkg/progress"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/client"
+	"github.com/spf13/cobra"
+)
+
+type pushOptions struct {
+	image    string
+	machines []string
+	platform string
+}
+
+func NewPushCommand() *cobra.Command {
+	opts := pushOptions{}
+
+	cmd := &cobra.Command{
+		Use:   "push IMAGE",
+		Short: "Upload a local Docker image to the cluster.",
+		Long: `Upload a local Docker image to the cluster transferring only the missing layers.
+The image is uploaded to all cluster machines (default) or the specified machine(s).`,
+		Example: `  # Push image to all machines in the cluster.
+  uc image push myapp:latest
+
+  # Push image to specific machine.
+  uc image push myapp:latest -m machine1
+
+  # Push image to multiple machines.
+  uc image push myapp:latest -m machine1,machine2,machine3
+
+  # Push a specific platform of a multi-platform image.
+  uc image push myapp:latest --platform linux/amd64`,
+		Args: cobra.ExactArgs(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+
+			opts.image = args[0]
+			return push(cmd.Context(), uncli, opts)
+		},
+	}
+
+	cmd.Flags().StringSliceVarP(&opts.machines, "machine", "m", nil,
+		"Machine names or IDs to push the image to. Can be specified multiple times or as a comma-separated list. "+
+			"(default is all machines)")
+	cmd.Flags().StringVar(
+		&opts.platform, "platform", "",
+		"Push a specific platform of a multi-platform image (e.g., linux/amd64, linux/arm64).\n"+
+			"Local Docker must be configured to use containerd image store to support multi-platform images.",
+	)
+
+	return cmd
+}
+
+func push(ctx context.Context, uncli *cli.CLI, opts pushOptions) error {
+	clusterClient, err := uncli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer clusterClient.Close()
+
+	machines := cli.ExpandCommaSeparatedValues(opts.machines)
+	pushOpts := client.PushImageOptions{}
+
+	// Special handling for an explicit "all" keyword to push to all machines.
+	if len(machines) == 1 && machines[0] == "all" {
+		pushOpts.AllMachines = true
+	} else if len(machines) > 0 {
+		pushOpts.Machines = machines
+	} else {
+		// Default is to push to all machines in the cluster.
+		pushOpts.AllMachines = true
+	}
+
+	if opts.platform != "" {
+		p, err := platforms.Parse(opts.platform)
+		if err != nil {
+			return fmt.Errorf("invalid platform '%s': %w", opts.platform, err)
+		}
+		pushOpts.Platform = &p
+	}
+
+	return progress.RunWithTitle(ctx, func(ctx context.Context) error {
+		if err = clusterClient.PushImage(ctx, opts.image, pushOpts); err != nil {
+			return fmt.Errorf("push image to cluster: %w", err)
+		}
+		return nil
+	}, uncli.ProgressOut(), fmt.Sprintf("Pushing image %s to cluster", opts.image))
+}

cmd/uncloud/image/root.go

@@ -0,0 +1,19 @@
+package image
+
+import (
+	"github.com/spf13/cobra"
+)
+
+func NewRootCommand() *cobra.Command {
+	cmd := &cobra.Command{
+		Use:   "image",
+		Short: "Manage images on machines in the cluster.",
+	}
+
+	cmd.AddCommand(
+		NewListCommand(),
+		NewPushCommand(),
+	)
+
+	return cmd
+}

cmd/uncloud/images.go

@@ -0,0 +1,19 @@
+package main
+
+import (
+	"strings"
+
+	"github.com/psviderski/uncloud/cmd/uncloud/image"
+	"github.com/spf13/cobra"
+)
+
+// NewImagesCommand returns the 'image ls' command modified to work as 'images'.
+func NewImagesCommand() *cobra.Command {
+	listCmd := image.NewListCommand()
+	listCmd.Use = "images [IMAGE]"
+	// Remove 'list' alias since this command is already an alias.
+	listCmd.Aliases = nil
+	listCmd.Example = strings.ReplaceAll(listCmd.Example, "uc image ls", "uc images")
+
+	return listCmd
+}

cmd/uncloud/machine/add.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"fmt"
 	"net/netip"
+	"strings"
 	"time"
 
 	"github.com/cenkalti/backoff/v4"
@@ -22,9 +23,9 @@ import (
 type addOptions struct {
 	name      string
 	noCaddy   bool
+	noInstall bool
 	publicIP  string
 	sshKey    string
-	context  string
 	version   string
 }
 
@@ -33,11 +34,22 @@ func NewAddCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:   "add [USER@]HOST[:PORT]",
 		Short: "Add a remote machine to a cluster.",
+		Long: `Add a new machine to an existing Uncloud cluster.
+
+Connection methods:
+  ssh://user@host       - Use built-in SSH library (default, no prefix required)
+  ssh+cli://user@host   - Use system SSH command (supports ProxyJump, SSH config)`,
 		Args: cobra.ExactArgs(1),
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
 
-			user, host, port, err := config.SSHDestination(args[0]).Parse()
+			// Determine if SSH CLI needs to be used and strip scheme
+			destination := args[0]
+			useSSHCLI := strings.HasPrefix(destination, "ssh+cli://")
+			destination = strings.TrimPrefix(destination, "ssh+cli://")
+			destination = strings.TrimPrefix(destination, "ssh://")
+
+			user, host, port, err := config.SSHDestination(destination).Parse()
 			if err != nil {
 				return fmt.Errorf("parse remote machine: %w", err)
 			}
@@ -46,6 +58,7 @@ func NewAddCommand() *cobra.Command {
 				Host:      host,
 				Port:      port,
 				KeyPath:   opts.sshKey,
+				UseSSHCLI: useSSHCLI,
 			}
 
 			return add(cmd.Context(), uncli, remoteMachine, opts)
@@ -56,6 +69,11 @@ func NewAddCommand() *cobra.Command {
 		&opts.noCaddy, "no-caddy", false,
 		"Don't deploy Caddy reverse proxy service to the machine.",
 	)
+	cmd.Flags().BoolVar(
+		&opts.noInstall, "no-install", false,
+		"Skip installation of Docker, Uncloud daemon, and dependencies on the machine. "+
+			"Assumes they're already installed and running.",
+	)
 	cmd.Flags().StringVar(
 		&opts.publicIP, "public-ip", "auto",
 		"Public IP address of the machine for ingress configuration. Use 'auto' for automatic detection, "+
@@ -70,10 +88,6 @@ func NewAddCommand() *cobra.Command {
 		&opts.version, "version", "latest",
 		"Version of the Uncloud daemon to install on the machine.",
 	)
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context to add the machine to. (default is the current context)",
-	)
 
 	return cmd
 }
@@ -94,10 +108,10 @@ func add(ctx context.Context, uncli *cli.CLI, remoteMachine *cli.RemoteMachine,
 	}
 
 	clusterClient, machineClient, err := uncli.AddMachine(ctx, cli.AddMachineOptions{
-		Context:       opts.context,
 		MachineName:   opts.name,
 		PublicIP:      publicIP,
 		RemoteMachine: remoteMachine,
+		SkipInstall:   opts.noInstall,
 		Version:       opts.version,
 	})
 	if err != nil {

cmd/uncloud/machine/init.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"fmt"
 	"net/netip"
+	"strings"
 
 	"github.com/docker/compose/v2/pkg/progress"
 	"github.com/psviderski/uncloud/cmd/uncloud/caddy"
@@ -22,6 +23,7 @@ type initOptions struct {
 	network     string
 	noCaddy     bool
 	noDNS       bool
+	noInstall   bool
 	publicIP    string
 	sshKey      string
 	version     string
@@ -31,9 +33,26 @@ type initOptions struct {
 func NewInitCommand() *cobra.Command {
 	opts := initOptions{}
 	cmd := &cobra.Command{
-		Use:   "init [USER@HOST:PORT]",
+		Use:   "init [schema://]USER@HOST[:PORT]",
 		Short: "Initialise a new cluster with a remote machine as the first member.",
-		// TODO: include usage examples of initialising a remote machine.
+		Long: `Initialise a new cluster by setting up a remote machine as the first member.
+This command creates a new context in your Uncloud config to manage the cluster.
+
+Connection methods:
+  ssh://user@host       - Use built-in SSH library (default, no prefix required)
+  ssh+cli://user@host   - Use system SSH command (supports ProxyJump, SSH config)`,
+		Example: `  # Initialise a new cluster with default settings.
+  uc machine init root@<your-server-ip>
+
+  # Initialise with a context name 'prod' in the Uncloud config (~/.config/uncloud/config.yaml) and machine name 'vps1'.
+  uc machine init root@<your-server-ip> -c prod -n vps1
+
+  # Initialise with a non-root user and custom SSH port and key.
+  uc machine init ubuntu@<your-server-ip>:2222 -i ~/.ssh/mykey
+
+  # Initialise without Caddy (no reverse proxy) and without an automatically managed domain name (xxxxxx.uncld.dev).
+  # You can deploy Caddy with 'uc caddy deploy' and reserve a domain with 'uc dns reserve' later.
+  uc machine init root@<your-server-ip> --no-caddy --no-dns`,
 		// TODO: support initialising a cluster on the local machine.
 		Args: cobra.MaximumNArgs(1),
 		RunE: func(cmd *cobra.Command, args []string) error {
@@ -41,7 +60,13 @@ func NewInitCommand() *cobra.Command {
 
 			var remoteMachine *cli.RemoteMachine
 			if len(args) > 0 {
-				user, host, port, err := config.SSHDestination(args[0]).Parse()
+				// Determine if SSH CLI is requested and strip scheme
+				destination := args[0]
+				useSSHCLI := strings.HasPrefix(destination, "ssh+cli://")
+				destination = strings.TrimPrefix(destination, "ssh+cli://")
+				destination = strings.TrimPrefix(destination, "ssh://")
+
+				user, host, port, err := config.SSHDestination(destination).Parse()
 				if err != nil {
 					return fmt.Errorf("parse remote machine: %w", err)
 				}
@@ -50,6 +75,7 @@ func NewInitCommand() *cobra.Command {
 					Host:      host,
 					Port:      port,
 					KeyPath:   opts.sshKey,
+					UseSSHCLI: useSSHCLI,
 				}
 			}
 
@@ -68,11 +94,16 @@ func NewInitCommand() *cobra.Command {
 	)
 	cmd.Flags().BoolVar(
 		&opts.noCaddy, "no-caddy", false,
-		"Don't deploy Caddy reverse proxy service to the machine.",
+		"Don't deploy Caddy reverse proxy service to the machine. You can deploy it later with 'uc caddy deploy'.",
 	)
 	cmd.Flags().BoolVar(
 		&opts.noDNS, "no-dns", false,
-		"Don't reserve a cluster domain in Uncloud DNS.",
+		"Don't reserve a cluster domain in Uncloud DNS. You can reserve it later with 'uc dns reserve'.",
+	)
+	cmd.Flags().BoolVar(
+		&opts.noInstall, "no-install", false,
+		"Skip installation of Docker, Uncloud daemon, and dependencies on the machine. "+
+			"Assumes they're already installed and running.",
 	)
 	cmd.Flags().StringVar(
 		&opts.publicIP, "public-ip", "auto",
@@ -90,13 +121,18 @@ func NewInitCommand() *cobra.Command {
 	)
 	cmd.Flags().StringVarP(
 		&opts.context, "context", "c", cli.DefaultContextName,
-		"Name of the created context for the initialised cluster in the Uncloud config.",
+		"Name of the new context to be created in the Uncloud config to manage the cluster.",
 	)
 
 	return cmd
 }
 
 func initCluster(ctx context.Context, uncli *cli.CLI, remoteMachine *cli.RemoteMachine, opts initOptions) error {
+	if uncli.Config == nil {
+		// Config is nil when connecting directly to a remote machine (--connect) without using Uncloud config.
+		return fmt.Errorf("do not specify --connect when initialising a new cluster")
+	}
+
 	netPrefix, err := netip.ParsePrefix(opts.network)
 	if err != nil {
 		return fmt.Errorf("parse network CIDR: %w", err)
@@ -121,6 +157,7 @@ func initCluster(ctx context.Context, uncli *cli.CLI, remoteMachine *cli.RemoteM
 		Network:       netPrefix,
 		PublicIP:      publicIP,
 		RemoteMachine: remoteMachine,
+		SkipInstall:   opts.noInstall,
 		Version:       opts.version,
 	})
 	if err != nil {

cmd/uncloud/machine/ls.go

@@ -14,25 +14,20 @@ import (
 )
 
 func NewListCommand() *cobra.Command {
-	var contextName string
 	cmd := &cobra.Command{
 		Use:     "ls",
 		Aliases: []string{"list"},
 		Short:   "List machines in a cluster.",
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-			return list(cmd.Context(), uncli, contextName)
+			return list(cmd.Context(), uncli)
 		},
 	}
-	cmd.Flags().StringVarP(
-		&contextName, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 	return cmd
 }
 
-func list(ctx context.Context, uncli *cli.CLI, clusterName string) error {
-	client, err := uncli.ConnectCluster(ctx, clusterName)
+func list(ctx context.Context, uncli *cli.CLI) error {
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
@@ -46,7 +41,7 @@ func list(ctx context.Context, uncli *cli.CLI, clusterName string) error {
 	// Print the list of machines in a table format.
 	tw := tabwriter.NewWriter(os.Stdout, 0, 0, 3, ' ', 0)
 	// Print header.
-	if _, err = fmt.Fprintln(tw, "NAME\tSTATE\tADDRESS\tPUBLIC IP\tWIREGUARD ENDPOINTS"); err != nil {
+	if _, err = fmt.Fprintln(tw, "NAME\tSTATE\tADDRESS\tPUBLIC IP\tWIREGUARD ENDPOINTS\tMACHINE ID"); err != nil {
 		return fmt.Errorf("write header: %w", err)
 	}
 	// Print rows.
@@ -68,8 +63,8 @@ func list(ctx context.Context, uncli *cli.CLI, clusterName string) error {
 		}
 
 		if _, err = fmt.Fprintf(
-			tw, "%s\t%s\t%s\t%s\t%s\n", m.Name, capitalise(member.State.String()), subnet, publicIP,
-			strings.Join(endpoints, ", "),
+			tw, "%s\t%s\t%s\t%s\t%s\t%s\n", m.Name, capitalise(member.State.String()), subnet, publicIP,
+			strings.Join(endpoints, ", "), member.Machine.Id,
 		); err != nil {
 			return fmt.Errorf("write row: %w", err)
 		}

cmd/uncloud/machine/rename.go

@@ -9,7 +9,6 @@ import (
 )
 
 func NewRenameCommand() *cobra.Command {
-	var contextName string
 	cmd := &cobra.Command{
 		Use:   "rename OLD_NAME NEW_NAME",
 		Short: "Rename a machine in the cluster.",
@@ -20,18 +19,14 @@ configuration including network settings, public IP, and cluster membership.`,
 		Args: cobra.ExactArgs(2),
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-			return rename(cmd.Context(), uncli, contextName, args[0], args[1])
+			return rename(cmd.Context(), uncli, args[0], args[1])
 		},
 	}
-	cmd.Flags().StringVarP(
-		&contextName, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 	return cmd
 }
 
-func rename(ctx context.Context, uncli *cli.CLI, contextName, oldName, newName string) error {
-	client, err := uncli.ConnectCluster(ctx, contextName)
+func rename(ctx context.Context, uncli *cli.CLI, oldName, newName string) error {
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return err
 	}

cmd/uncloud/machine/rm.go

@@ -22,7 +22,6 @@ import (
 type removeOptions struct {
 	noReset bool
 	yes     bool
-	context string
 }
 
 func NewRmCommand() *cobra.Command {
@@ -39,8 +38,6 @@ func NewRmCommand() *cobra.Command {
 		},
 	}
 
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
 	cmd.Flags().BoolVarP(&opts.yes, "yes", "y", false,
 		"Do not prompt for confirmation before removing the machine.")
 	cmd.Flags().BoolVar(&opts.noReset, "no-reset", false,
@@ -51,14 +48,14 @@ func NewRmCommand() *cobra.Command {
 
 func remove(ctx context.Context, uncli *cli.CLI, nameOrID string, opts removeOptions) error {
 	// TODO: automatically choose a connection to the machine that is not being removed.
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
 	defer client.Close()
 
 	// Verify the machine exists and list all service containers on it including stopped ones.
-	mctx, machines, err := api.ProxyMachinesContext(ctx, client, []string{nameOrID})
+	mctx, machines, err := client.ProxyMachinesContext(ctx, []string{nameOrID})
 	if err != nil {
 		return err
 	}
@@ -155,8 +152,21 @@ func remove(ctx context.Context, uncli *cli.CLI, nameOrID string, opts removeOpt
 		}
 	}
 
-	// TODO: remove the connection to the machine from the uncloud config if it exists. We need a way to associate
-	//  the machine with its connection in the config, e.g. by storing the machine name in the connection metadata.
+	// Remove the connection to the machine from the uncloud config if it exists.
+	if uncli.Config != nil {
+		contextName := uncli.GetContextOverrideOrCurrent()
+		if context, ok := uncli.Config.Contexts[contextName]; ok {
+			for i, c := range context.Connections {
+				if c.MachineID == m.Id {
+					context.Connections = slices.Delete(context.Connections, i, i+1)
+					break
+				}
+			}
+			if err := uncli.Config.Save(); err != nil {
+				return fmt.Errorf("save config: %w", err)
+			}
+		}
+	}
 
 	// TODO: If Caddy was running on this machine and a cluster domain is reserved,
 	//  let the user know that the DNS records should be updated.

cmd/uncloud/machine/root.go

@@ -8,7 +8,7 @@ func NewRootCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:     "machine",
 		Aliases: []string{"m"},
-		Short:   "Manage machines in an Uncloud cluster.",
+		Short:   "Manage machines in the cluster.",
 	}
 	cmd.AddCommand(
 		NewAddCommand(),

cmd/uncloud/machine/update.go

@@ -13,7 +13,6 @@ import (
 type updateOptions struct {
 	name     string
 	publicIP string
-	context  string
 }
 
 func NewUpdateCommand() *cobra.Command {
@@ -43,10 +42,6 @@ At least one flag must be specified to perform an update operation.`,
 		&opts.publicIP, "public-ip", "",
 		fmt.Sprintf("Public IP address of the machine for ingress configuration. Use '%s' or '' to remove the public IP.", PublicIPNone),
 	)
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 
 	return cmd
 }
@@ -57,7 +52,7 @@ func update(ctx context.Context, uncli *cli.CLI, cmd *cobra.Command, opts update
 		return fmt.Errorf("at least one update flag must be specified (--name, --public-ip)")
 	}
 
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return err
 	}

cmd/uncloud/main.go

@@ -6,15 +6,18 @@ import (
 	"net/netip"
 	"strings"
 
+	"github.com/charmbracelet/lipgloss"
 	"github.com/psviderski/uncloud/cmd/uncloud/caddy"
 	cmdcontext "github.com/psviderski/uncloud/cmd/uncloud/context"
 	"github.com/psviderski/uncloud/cmd/uncloud/dns"
+	"github.com/psviderski/uncloud/cmd/uncloud/image"
 	"github.com/psviderski/uncloud/cmd/uncloud/machine"
 	"github.com/psviderski/uncloud/cmd/uncloud/service"
 	"github.com/psviderski/uncloud/cmd/uncloud/volume"
 	"github.com/psviderski/uncloud/internal/cli"
 	"github.com/psviderski/uncloud/internal/cli/config"
 	"github.com/psviderski/uncloud/internal/fs"
+	"github.com/psviderski/uncloud/internal/log"
 	"github.com/psviderski/uncloud/internal/version"
 	"github.com/spf13/cobra"
 )
@@ -22,17 +25,24 @@ import (
 type globalOptions struct {
 	configPath string
 	connect    string
+	context    string
 }
 
 func main() {
+	log.InitLoggerFromEnv()
+
 	opts := globalOptions{}
 	cmd := &cobra.Command{
 		Use:           "uc",
-		Short:         "A CLI tool for managing Uncloud resources such as clusters, machines, and services.",
+		Short:         "A CLI tool for managing Uncloud resources such as machines, services, and volumes.",
 		Version:       version.String(),
 		SilenceUsage:  true,
 		SilenceErrors: true,
 		PersistentPreRunE: func(cmd *cobra.Command, args []string) error {
+			cli.BindEnvToFlag(cmd, "connect", "UNCLOUD_CONNECT")
+			cli.BindEnvToFlag(cmd, "context", "UNCLOUD_CONTEXT")
+			cli.BindEnvToFlag(cmd, "uncloud-config", "UNCLOUD_CONFIG")
+
 			var conn *config.MachineConnection
 			if opts.connect != "" {
 				if strings.HasPrefix(opts.connect, "tcp://") {
@@ -43,11 +53,17 @@ func main() {
 					conn = &config.MachineConnection{
 						TCP: &addrPort,
 					}
-				} else {
-					dest := opts.connect
-					if strings.HasPrefix(dest, "ssh://") {
-						dest = dest[len("ssh://"):]
+				} else if strings.HasPrefix(opts.connect, "ssh+cli://") {
+					dest := opts.connect[len("ssh+cli://"):]
+					conn = &config.MachineConnection{
+						SSHCLI: config.SSHDestination(dest),
 					}
+				} else if strings.HasPrefix(opts.connect, "unix://") {
+					conn = &config.MachineConnection{
+						Unix: opts.connect[len("unix://"):],
+					}
+				} else {
+					dest := strings.TrimPrefix(opts.connect, "ssh://")
 					conn = &config.MachineConnection{
 						SSH: config.SSHDestination(dest),
 					}
@@ -55,7 +71,7 @@ func main() {
 			}
 
 			configPath := fs.ExpandHomeDir(opts.configPath)
-			uncli, err := cli.New(configPath, conn)
+			uncli, err := cli.New(configPath, conn, opts.context)
 			if err != nil {
 				return fmt.Errorf("initialise CLI: %w", err)
 			}
@@ -65,28 +81,58 @@ func main() {
 	}
 
 	cmd.PersistentFlags().StringVar(&opts.connect, "connect", "",
-		"Connect to a remote cluster machine without using the Uncloud configuration file.\n"+
-			"Format: [ssh://]user@host[:port] or tcp://host:port")
-	// TODO: allow to override using UNCLOUD_CONFIG env var.
+		"Connect to a remote cluster machine without using the Uncloud configuration file. [$UNCLOUD_CONNECT]\n"+
+			"Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], tcp://host:port, or unix:///path/to/uncloud.sock")
 	cmd.PersistentFlags().StringVar(&opts.configPath, "uncloud-config", "~/.config/uncloud/config.yaml",
-		"Path to the Uncloud configuration file.")
+		"Path to the Uncloud configuration file. [$UNCLOUD_CONFIG]")
 	_ = cmd.MarkPersistentFlagFilename("uncloud-config", "yaml", "yml")
-	// TODO: make --context a global flag and pass it as a value of the command context.
+	cmd.PersistentFlags().StringVarP(&opts.context, "context", "c", "",
+		"Name of the cluster context to use (default is the current context). [$UNCLOUD_CONTEXT]")
+
+	// Set custom help function to show links to docs and Discord only for the root 'uc' command.
+	defaultHelpFunc := cmd.HelpFunc()
+	cmd.SetHelpFunc(func(c *cobra.Command, args []string) {
+		defaultHelpFunc(c, args)
+		// Only show links for the root 'uc' command.
+		if c.Name() == "uc" {
+			urlStyle := lipgloss.NewStyle().
+				Underline(true).
+				Foreground(lipgloss.Color("12")) // light blue
+
+			fmt.Fprintln(c.OutOrStdout())
+			fmt.Fprintf(c.OutOrStdout(), "Learn more about Uncloud:       %s\n",
+				urlStyle.Render("https://uncloud.run/docs"))
+			fmt.Fprintf(c.OutOrStdout(), "Join our Discord community:     %s\n",
+				urlStyle.Render("https://uncloud.run/discord"))
+		}
+	})
+
+	cmd.AddGroup(&cobra.Group{
+		ID:    "service",
+		Title: "Deploy and manage services:",
+	})
 
 	cmd.AddCommand(
+		NewBuildCommand(),
 		NewDeployCommand(),
 		NewDocsCommand(),
-		NewBuildCommand(),
+		NewImagesCommand(),
+		NewPsCommand(),
 		caddy.NewRootCommand(),
 		cmdcontext.NewRootCommand(),
 		dns.NewRootCommand(),
+		image.NewRootCommand(),
 		machine.NewRootCommand(),
 		service.NewRootCommand(),
-		service.NewInspectCommand(),
-		service.NewListCommand(),
-		service.NewRmCommand(),
-		service.NewRunCommand(),
-		service.NewScaleCommand(),
+		service.NewExecCommand("service"),
+		service.NewInspectCommand("service"),
+		service.NewListCommand("service"),
+		service.NewLogsCommand("service"),
+		service.NewRmCommand("service"),
+		service.NewRunCommand("service"),
+		service.NewScaleCommand("service"),
+		service.NewStartCommand("service"),
+		service.NewStopCommand("service"),
 		volume.NewRootCommand(),
 	)
 	cobra.CheckErr(cmd.Execute())

cmd/uncloud/ps.go

@@ -0,0 +1,284 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"sort"
+	"time"
+
+	"github.com/charmbracelet/huh/spinner"
+	"github.com/charmbracelet/lipgloss"
+	"github.com/charmbracelet/lipgloss/table"
+	"github.com/docker/docker/api/types/container"
+	"github.com/docker/go-units"
+	"github.com/spf13/cobra"
+
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/client"
+)
+
+const (
+	sortByService = "service"
+	sortByMachine = "machine"
+	sortByHealth  = "health"
+)
+
+type containerHighlight int
+
+const (
+	highlightDanger containerHighlight = iota
+	highlightWarning
+	highlightSuccess
+	highlightNormal
+)
+
+type psOptions struct {
+	sortBy string
+}
+
+func NewPsCommand() *cobra.Command {
+	opts := psOptions{}
+	cmd := &cobra.Command{
+		Use:   "ps",
+		Short: "List all service containers.",
+		Long: `List all service containers across all machines in the cluster.
+
+This command provides a comprehensive overview of all running containers that are part of a service,
+making it easy to see the distribution and status of containers across the cluster.`,
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+
+			if opts.sortBy != sortByService && opts.sortBy != sortByMachine && opts.sortBy != sortByHealth {
+				return fmt.Errorf("invalid value for --sort: %q, must be one of '%s', '%s' or '%s'", opts.sortBy,
+					sortByService, sortByMachine, sortByHealth)
+			}
+
+			return runPs(cmd.Context(), uncli, opts)
+		},
+		GroupID: "service",
+	}
+	cmd.Flags().StringVarP(&opts.sortBy, "sort", "s", sortByService,
+		"Sort containers by 'service', 'machine', or 'health'.")
+	return cmd
+}
+
+type containerInfo struct {
+	serviceName string
+	machineName string
+	id          string
+	name        string
+	image       string
+	status      string
+	highlight   containerHighlight
+	created     time.Time
+	ip          string
+}
+
+func runPs(ctx context.Context, uncli *cli.CLI, opts psOptions) error {
+	clusterClient, err := uncli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer clusterClient.Close()
+
+	var containers []containerInfo
+	err = spinner.New().
+		Title(" Collecting container info...").
+		Type(spinner.MiniDot).
+		Style(lipgloss.NewStyle().Foreground(lipgloss.Color("3"))).
+		ActionWithErr(func(ctx context.Context) error {
+			containers, err = collectContainers(ctx, clusterClient)
+			return err
+		}).
+		Run()
+	if err != nil {
+		return fmt.Errorf("collect containers: %w", err)
+	}
+
+	// Sort the containers based on the sorting option.
+	sort.SliceStable(containers, func(i, j int) bool {
+		a, b := containers[i], containers[j]
+		switch opts.sortBy {
+		case sortByHealth:
+			if a.highlight != b.highlight {
+				return a.highlight < b.highlight
+			}
+			if a.serviceName != b.serviceName {
+				return a.serviceName < b.serviceName
+			}
+		case sortByMachine:
+			if a.machineName != b.machineName {
+				return a.machineName < b.machineName
+			}
+			if a.serviceName != b.serviceName {
+				return a.serviceName < b.serviceName
+			}
+		default: // sortByService
+			if a.serviceName != b.serviceName {
+				return a.serviceName < b.serviceName
+			}
+		}
+		// Fallback to creation time (newest first).
+		return a.created.After(b.created)
+	})
+
+	return printContainers(containers)
+}
+
+func printContainers(containers []containerInfo) error {
+	t := table.New().
+		// Remove the default border.
+		Border(lipgloss.Border{}).
+		BorderTop(false).
+		BorderBottom(false).
+		BorderLeft(false).
+		BorderRight(false).
+		BorderHeader(false).
+		BorderColumn(false).
+		StyleFunc(func(row, col int) lipgloss.Style {
+			if row == table.HeaderRow {
+				return lipgloss.NewStyle().Bold(true).PaddingRight(3)
+			}
+			// Regular style for data rows with padding.
+			return lipgloss.NewStyle().PaddingRight(3)
+		})
+
+	t.Headers("SERVICE", "CONTAINER ID", "CONTAINER NAME", "IMAGE", "CREATED", "STATUS", "IP ADDRESS", "MACHINE")
+
+	for _, ctr := range containers {
+		id := ctr.id
+		if len(id) > 12 {
+			id = id[:12]
+		}
+
+		created := units.HumanDuration(time.Now().UTC().Sub(ctr.created)) + " ago"
+
+		var statusStyle lipgloss.Style
+		switch ctr.highlight {
+		case highlightSuccess:
+			statusStyle = lipgloss.NewStyle().Foreground(lipgloss.Color("2")) // Green
+		case highlightDanger:
+			statusStyle = lipgloss.NewStyle().Foreground(lipgloss.Color("1")) // Red
+		case highlightWarning:
+			statusStyle = lipgloss.NewStyle().Foreground(lipgloss.Color("3")) // Yellow
+		default:
+			statusStyle = lipgloss.NewStyle() // Default
+		}
+
+		t.Row(
+			ctr.serviceName,
+			id,
+			ctr.name,
+			ctr.image,
+			created,
+			statusStyle.Render(ctr.status),
+			ctr.ip,
+			ctr.machineName,
+		)
+	}
+
+	fmt.Println(t)
+	return nil
+}
+
+func collectContainers(ctx context.Context, cli *client.Client) ([]containerInfo, error) {
+	listCtx, machines, err := cli.ProxyMachinesContext(ctx, nil)
+	if err != nil {
+		return nil, fmt.Errorf("proxy machines context: %w", err)
+	}
+
+	// Create a map of IP to machine name for resolving response metadata
+	machinesNamesByIP := make(map[string]string)
+	for _, m := range machines {
+		if addr, err := m.Machine.Network.ManagementIp.ToAddr(); err == nil {
+			machinesNamesByIP[addr.String()] = m.Machine.Name
+		}
+	}
+
+	// List all service containers across all machines in the cluster.
+	machineContainers, err := cli.Docker.ListServiceContainers(
+		listCtx, "", container.ListOptions{All: true},
+	)
+	if err != nil {
+		return nil, fmt.Errorf("list service containers: %w", err)
+	}
+
+	var containers []containerInfo
+	for _, msc := range machineContainers {
+		machineName := "unknown"
+
+		// Metadata can be nil if the request was broadcasted to only one machine.
+		if msc.Metadata == nil && len(machineContainers) > 1 {
+			return nil, fmt.Errorf("something went wrong with gRPC proxy: metadata is missing for a machine response")
+		}
+		if msc.Metadata != nil && msc.Metadata.Error != "" {
+			client.PrintWarning(fmt.Sprintf("failed to list containers on machine %s: %s", machineName,
+				msc.Metadata.Error))
+			continue
+		}
+
+		if msc.Metadata != nil {
+			var ok bool
+			machineName, ok = machinesNamesByIP[msc.Metadata.Machine]
+			if !ok {
+				// Fallback to machine's IP as name.
+				machineName = msc.Metadata.Machine
+			}
+		} else {
+			// Fallback to the first available machine name.
+			if len(machines) > 0 {
+				machineName = machines[0].Machine.Name
+			}
+		}
+
+		for _, ctr := range msc.Containers {
+			if ctr.Container.State == nil || ctr.Container.Config == nil {
+				continue
+			}
+
+			status, err := ctr.Container.HumanState()
+			if err != nil {
+				return nil, fmt.Errorf("get human state for container %s: %w", ctr.Container.ID, err)
+			}
+
+			var highlight containerHighlight
+			healthStatus := ""
+			if ctr.Container.State.Health != nil {
+				healthStatus = ctr.Container.State.Health.Status
+			}
+
+			if healthStatus == container.Unhealthy || ctr.Container.State.Status == "dead" || ctr.Container.State.OOMKilled || ctr.Container.State.Dead {
+				highlight = highlightDanger
+			} else if healthStatus == container.Healthy {
+				highlight = highlightSuccess
+			} else if ctr.Container.State.Status == "running" {
+				highlight = highlightNormal
+			} else { // Other non-critical but noteworthy states
+				highlight = highlightWarning
+			}
+
+			created, _ := time.Parse(time.RFC3339Nano, ctr.Container.Created)
+
+			ip := ctr.Container.UncloudNetworkIP()
+			ipStr := ""
+			// The container might not have an IP if it's not running or uses the host network.
+			if ip.IsValid() {
+				ipStr = ip.String()
+			}
+
+			info := containerInfo{
+				serviceName: ctr.ServiceName(),
+				machineName: machineName,
+				id:          ctr.Container.ID,
+				name:        ctr.Container.Name,
+				image:       ctr.Container.Config.Image,
+				status:      status,
+				highlight:   highlight,
+				created:     created,
+				ip:          ipStr,
+			}
+			containers = append(containers, info)
+		}
+	}
+	return containers, nil
+}

cmd/uncloud/ps_test.go

@@ -0,0 +1,442 @@
+package main
+
+import (
+	"context"
+	"encoding/json"
+	"net/netip"
+	"testing"
+
+	"github.com/psviderski/uncloud/internal/machine/api/pb"
+	"github.com/psviderski/uncloud/internal/machine/docker"
+	"github.com/psviderski/uncloud/pkg/client"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"google.golang.org/grpc"
+	"google.golang.org/protobuf/types/known/emptypb"
+)
+
+// mockDockerClient implements pb.DockerClient
+type mockDockerClient struct {
+	pb.DockerClient // Embed to avoid implementing all methods
+	listResp        *pb.ListServiceContainersResponse
+	listErr         error
+}
+
+func (m *mockDockerClient) ListServiceContainers(ctx context.Context, in *pb.ListServiceContainersRequest, opts ...grpc.CallOption) (*pb.ListServiceContainersResponse, error) {
+	return m.listResp, m.listErr
+}
+
+// mockClusterClient implements pb.ClusterClient
+type mockClusterClient struct {
+	pb.ClusterClient // Embed to avoid implementing all methods
+	machinesResp     *pb.ListMachinesResponse
+	machinesErr      error
+}
+
+func (m *mockClusterClient) ListMachines(ctx context.Context, in *emptypb.Empty, opts ...grpc.CallOption) (*pb.ListMachinesResponse, error) {
+	return m.machinesResp, m.machinesErr
+}
+
+func TestCollectContainers_NilMetadata(t *testing.T) {
+	// Setup container data
+	containerData := map[string]interface{}{
+		"Id":   "container1",
+		"Name": "test-container",
+		"Config": map[string]interface{}{
+			"Image": "test-image",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON, err := json.Marshal(containerData)
+	require.NoError(t, err)
+
+	serviceSpecJSON, err := json.Marshal(map[string]interface{}{})
+	require.NoError(t, err)
+
+	// Setup mocks
+	mockDocker := &mockDockerClient{
+		listResp: &pb.ListServiceContainersResponse{
+			Messages: []*pb.MachineServiceContainers{
+				{
+					Metadata: nil, // Simulating the issue: nil metadata
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	machineIP := "10.0.0.1"
+	mockCluster := &mockClusterClient{
+		machinesResp: &pb.ListMachinesResponse{
+			Machines: []*pb.MachineMember{
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-1",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr(machineIP)),
+						},
+					},
+					State: pb.MachineMember_UP,
+				},
+			},
+		},
+	}
+
+	// Construct client with mocks
+	cli := &client.Client{
+		Docker: &docker.Client{
+			GRPCClient: mockDocker,
+		},
+		ClusterClient: mockCluster,
+	}
+
+	// Execute
+	containers, err := collectContainers(context.Background(), cli)
+	require.NoError(t, err)
+
+	// Verify
+	assert.Len(t, containers, 1)
+	if len(containers) > 0 {
+		c := containers[0]
+		assert.Equal(t, "container1", c.id)
+		assert.Equal(t, "test-container", c.name)
+		assert.Equal(t, "machine-1", c.machineName, "Should fall back to the single machine name when metadata is nil")
+	}
+}
+
+func TestCollectContainers_NilMetadata_MultipleMachines_Error(t *testing.T) {
+	// If we have multiple machines but receive nil metadata, it should return an error as it is ambiguous
+
+	// Setup container data
+	containerData1 := map[string]interface{}{
+		"Id": "container1",
+	}
+	containerJSON1, _ := json.Marshal(containerData1)
+
+	containerData2 := map[string]interface{}{
+		"Id": "container2",
+		"Config": map[string]interface{}{
+			"Image": "test-image",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON2, _ := json.Marshal(containerData2)
+
+	serviceSpecJSON, _ := json.Marshal(map[string]interface{}{})
+
+	// Setup mocks
+	mockDocker := &mockDockerClient{
+		listResp: &pb.ListServiceContainersResponse{
+			Messages: []*pb.MachineServiceContainers{
+				{
+					Metadata: nil, // Nil metadata
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON1,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+				{
+					Metadata: &pb.Metadata{Machine: "10.0.0.2"},
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON2,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	mockCluster := &mockClusterClient{
+		machinesResp: &pb.ListMachinesResponse{
+			Machines: []*pb.MachineMember{
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-1",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr("10.0.0.1")),
+						},
+					},
+				},
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-2",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr("10.0.0.2")),
+						},
+					},
+				},
+			},
+		},
+	}
+
+	// Construct client with mocks
+	cli := &client.Client{
+		Docker: &docker.Client{
+			GRPCClient: mockDocker,
+		},
+		ClusterClient: mockCluster,
+	}
+
+	// Execute
+	_, err := collectContainers(context.Background(), cli)
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "metadata is missing for a machine response")
+}
+
+func TestCollectContainers_MetadataPresent_MultipleMachines(t *testing.T) {
+	// Verify correct mapping of containers to machines when metadata is present
+
+	// Setup container data
+	containerData1 := map[string]interface{}{
+		"Id":   "container1",
+		"Name": "container-1",
+		"Config": map[string]interface{}{
+			"Image": "image-1",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON1, _ := json.Marshal(containerData1)
+
+	containerData2 := map[string]interface{}{
+		"Id":   "container2",
+		"Name": "container-2",
+		"Config": map[string]interface{}{
+			"Image": "image-2",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON2, _ := json.Marshal(containerData2)
+
+	serviceSpecJSON, _ := json.Marshal(map[string]interface{}{})
+
+	// Setup mocks
+	mockDocker := &mockDockerClient{
+		listResp: &pb.ListServiceContainersResponse{
+			Messages: []*pb.MachineServiceContainers{
+				{
+					Metadata: &pb.Metadata{Machine: "10.0.0.1"},
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON1,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+				{
+					Metadata: &pb.Metadata{Machine: "10.0.0.2"},
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON2,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	mockCluster := &mockClusterClient{
+		machinesResp: &pb.ListMachinesResponse{
+			Machines: []*pb.MachineMember{
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-1",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr("10.0.0.1")),
+						},
+					},
+				},
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-2",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr("10.0.0.2")),
+						},
+					},
+				},
+			},
+		},
+	}
+
+	cli := &client.Client{
+		Docker:        &docker.Client{GRPCClient: mockDocker},
+		ClusterClient: mockCluster,
+	}
+
+	containers, err := collectContainers(context.Background(), cli)
+	require.NoError(t, err)
+
+	assert.Len(t, containers, 2)
+	// Order is not guaranteed by the map iteration in logic or parallel fetch (though here it's mocked sequential),
+	// but collectContainers just appends.
+	// We'll find them by ID.
+	for _, c := range containers {
+		if c.id == "container1" {
+			assert.Equal(t, "machine-1", c.machineName)
+		} else if c.id == "container2" {
+			assert.Equal(t, "machine-2", c.machineName)
+		} else {
+			t.Errorf("unexpected container id: %s", c.id)
+		}
+	}
+}
+
+func TestCollectContainers_NilMetadata_NoMachines(t *testing.T) {
+	// Case: 1 msc with nil metadata but no machines at all
+
+	containerData := map[string]interface{}{
+		"Id": "container1",
+		"Config": map[string]interface{}{
+			"Image": "test-image",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON, _ := json.Marshal(containerData)
+	serviceSpecJSON, _ := json.Marshal(map[string]interface{}{})
+
+	mockDocker := &mockDockerClient{
+		listResp: &pb.ListServiceContainersResponse{
+			Messages: []*pb.MachineServiceContainers{
+				{
+					Metadata: nil,
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	// No machines in cluster response
+	mockCluster := &mockClusterClient{
+		machinesResp: &pb.ListMachinesResponse{
+			Machines: []*pb.MachineMember{},
+		},
+	}
+
+	cli := &client.Client{
+		Docker:        &docker.Client{GRPCClient: mockDocker},
+		ClusterClient: mockCluster,
+	}
+
+	containers, err := collectContainers(context.Background(), cli)
+	require.NoError(t, err)
+
+	assert.Len(t, containers, 1)
+	if len(containers) > 0 {
+		assert.Equal(t, "unknown", containers[0].machineName)
+	}
+}
+
+func TestCollectContainers_MetadataPresent_NotInMapping(t *testing.T) {
+	// Case: msc with metadata that is not in the IP-to-name mapping
+
+	containerData := map[string]interface{}{
+		"Id": "container1",
+		"Config": map[string]interface{}{
+			"Image": "test-image",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON, _ := json.Marshal(containerData)
+	serviceSpecJSON, _ := json.Marshal(map[string]interface{}{})
+
+	mockDocker := &mockDockerClient{
+		listResp: &pb.ListServiceContainersResponse{
+			Messages: []*pb.MachineServiceContainers{
+				{
+					Metadata: &pb.Metadata{Machine: "10.0.0.99"}, // Unknown IP
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	mockCluster := &mockClusterClient{
+		machinesResp: &pb.ListMachinesResponse{
+			Machines: []*pb.MachineMember{
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-1",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr("10.0.0.1")),
+						},
+					},
+				},
+			},
+		},
+	}
+
+	cli := &client.Client{
+		Docker:        &docker.Client{GRPCClient: mockDocker},
+		ClusterClient: mockCluster,
+	}
+
+	containers, err := collectContainers(context.Background(), cli)
+	require.NoError(t, err)
+
+	assert.Len(t, containers, 1)
+	if len(containers) > 0 {
+		// Should fallback to the IP/string in metadata
+		assert.Equal(t, "10.0.0.99", containers[0].machineName)
+	}
+}

cmd/uncloud/service/exec.go

@@ -0,0 +1,128 @@
+package service
+
+import (
+	"context"
+	"fmt"
+	"os"
+
+	"github.com/docker/cli/cli/streams"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/api"
+	"github.com/spf13/cobra"
+)
+
+type execCliOptions struct {
+	detach      bool
+	interactive bool
+	noTty       bool
+	containerId string
+}
+
+var DEFAULT_COMMAND = []string{"sh", "-c", "command -v bash >/dev/null 2>&1 && exec bash || exec sh"}
+
+func NewExecCommand(groupID string) *cobra.Command {
+	opts := execCliOptions{}
+
+	execCmd := &cobra.Command{
+		Use:   "exec [OPTIONS] SERVICE [COMMAND ARGS...]",
+		Short: "Execute a command in a running service container.",
+		Long: `Execute a command (interactive shell by default) in a running container within a service.
+If the service has multiple replicas and no container ID is specified, the command will be executed in a random container.
+	`,
+		Example: `
+  # Start an interactive shell ("bash" or "sh" will be tried by default)
+  uc exec web-service
+
+  # Start an interactive shell with explicit command
+  uc exec web-service /bin/zsh
+
+  # List files in the specific container of the service; --container accepts full ID or a (unique) prefix
+  uc exec --container d792e web-service ls -la
+
+  # Pipe input to a command inside the service container
+  cat backup.sql | uc exec -T db-service psql -U postgres mydb
+
+  # Run a task in the background (detached mode)
+  uc exec -d web-service /scripts/cleanup.sh`,
+		Args: cobra.MinimumNArgs(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			serviceName := args[0]
+			command := args[1:]
+			if len(command) == 0 {
+				command = DEFAULT_COMMAND
+			}
+			return runExec(cmd.Context(), uncli, serviceName, command, opts)
+		},
+		GroupID: groupID,
+	}
+
+	execCmd.Flags().BoolVarP(&opts.detach, "detach", "d", false, "Detached mode: run command in the background")
+
+	execCmd.Flags().BoolVarP(&opts.noTty, "no-tty", "T", !cli.IsStdoutTerminal(),
+		"Disable pseudo-TTY allocation. By default 'uc exec' allocates a TTY when connected to a terminal.")
+
+	// Keep "-i" and "-t" flags hidden for compatibility with docker exec
+	execCmd.Flags().BoolVarP(&opts.interactive, "interactive", "i", true, "Keep STDIN open even if not attached")
+	execCmd.Flags().MarkHidden("interactive")
+
+	execCmd.Flags().BoolP("tty", "t", false, "Allocate a pseudo-TTY")
+	execCmd.Flags().MarkHidden("tty")
+
+	// Common flags
+	execCmd.Flags().StringVar(&opts.containerId, "container", "",
+		"ID of the container to exec into. Accepts full ID or a unique prefix "+
+			"(default is the random container of the service)")
+
+	// This tells Cobra that all flags must come before positional arguments, so that
+	// commands with their own flags can be handled correctly.
+	execCmd.Flags().SetInterspersed(false)
+
+	return execCmd
+}
+
+func runExec(ctx context.Context, uncli *cli.CLI, serviceName string, command []string, opts execCliOptions) error {
+	if !opts.detach {
+		// Check if we're trying to attach to a TTY from a non-TTY client, e.g.
+		// when doing an 'cmd | uc exec ...'
+		stdin := streams.NewIn(os.Stdin)
+		// TODO: this logic/behavior mirrors docker-compose, but we can be smarter about it and detect TTY dynamically
+		if err := stdin.CheckTty(opts.interactive, !opts.noTty); err != nil {
+			return fmt.Errorf("check TTY: %w; use -T option to disable TTY allocation", err)
+		}
+	}
+
+	client, err := uncli.ConnectClusterWithOptions(ctx, cli.ConnectOptions{
+		ShowProgress: false,
+	})
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer client.Close()
+
+	execConfig := api.ExecOptions{
+		Command:     command,
+		AttachStdin: opts.interactive,
+		Tty:         !opts.noTty,
+		Detach:      opts.detach,
+	}
+
+	if !opts.detach {
+		execConfig.AttachStdout = true
+		execConfig.AttachStderr = true
+	}
+
+	exitCode, err := client.ExecContainer(ctx, serviceName, opts.containerId, execConfig)
+	if err != nil {
+		return fmt.Errorf("exec container: %w", err)
+	}
+
+	// For non-detached mode, exit with the same code as the executed command
+	if !opts.detach {
+		if exitCode != 0 {
+			os.Exit(exitCode)
+		}
+	}
+
+	return nil
+}

cmd/uncloud/service/inspect.go

@@ -4,21 +4,22 @@ import (
 	"context"
 	"fmt"
 	"os"
+	"slices"
 	"text/tabwriter"
 	"time"
 
 	"github.com/docker/docker/pkg/stringid"
 	"github.com/docker/go-units"
 	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/api"
 	"github.com/spf13/cobra"
 )
 
 type inspectOptions struct {
 	service string
-	cluster string
 }
 
-func NewInspectCommand() *cobra.Command {
+func NewInspectCommand(groupID string) *cobra.Command {
 	opts := inspectOptions{}
 	cmd := &cobra.Command{
 		Use:   "inspect SERVICE",
@@ -29,16 +30,13 @@ func NewInspectCommand() *cobra.Command {
 			opts.service = args[0]
 			return inspect(cmd.Context(), uncli, opts)
 		},
+		GroupID: groupID,
 	}
-	cmd.Flags().StringVarP(
-		&opts.cluster, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 	return cmd
 }
 
 func inspect(ctx context.Context, uncli *cli.CLI, opts inspectOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.cluster)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
@@ -58,23 +56,31 @@ func inspect(ctx context.Context, uncli *cli.CLI, opts inspectOptions) error {
 		machinesNamesByID[m.Machine.Id] = m.Machine.Name
 	}
 
-	fmt.Printf("ID:    %s\n", svc.ID)
+	fmt.Printf("Service ID: %s\n", svc.ID)
 	fmt.Printf("Name:       %s\n", svc.Name)
 	fmt.Printf("Mode:       %s\n", svc.Mode)
 	fmt.Println()
 
+	// Parse created times for sorting and display.
+	createdTimes := make(map[string]time.Time, len(svc.Containers))
+	for _, ctr := range svc.Containers {
+		createdTimes[ctr.Container.ID], _ = time.Parse(time.RFC3339Nano, ctr.Container.Created)
+	}
+
+	// Sort containers by created time (newest first).
+	slices.SortFunc(svc.Containers, func(a, b api.MachineServiceContainer) int {
+		return createdTimes[b.Container.ID].Compare(createdTimes[a.Container.ID])
+	})
+
 	// Print the list of containers in a table format.
 	tw := tabwriter.NewWriter(os.Stdout, 0, 0, 3, ' ', 0)
-	if _, err = fmt.Fprintln(tw, "CONTAINER ID\tIMAGE\tCREATED\tSTATUS\tMACHINE"); err != nil {
+	if _, err = fmt.Fprintln(tw, "CONTAINER ID\tIMAGE\tCREATED\tSTATUS\tIP ADDRESS\tMACHINE"); err != nil {
 		return fmt.Errorf("write header: %w", err)
 	}
 
+	now := time.Now().UTC()
 	for _, ctr := range svc.Containers {
-		createdAt, err := time.Parse(time.RFC3339Nano, ctr.Container.Created)
-		if err != nil {
-			return fmt.Errorf("parse created time: %w", err)
-		}
-		created := units.HumanDuration(time.Now().UTC().Sub(createdAt)) + " ago"
+		created := units.HumanDuration(now.Sub(createdTimes[ctr.Container.ID])) + " ago"
 
 		machine := machinesNamesByID[ctr.MachineID]
 		if machine == "" {
@@ -85,13 +91,21 @@ func inspect(ctx context.Context, uncli *cli.CLI, opts inspectOptions) error {
 			return fmt.Errorf("get human state: %w", err)
 		}
 
+		ip := ctr.Container.UncloudNetworkIP()
+		ipStr := ""
+		// The container might not have an IP if it's not running or uses the host network.
+		if ip.IsValid() {
+			ipStr = ip.String()
+		}
+
 		_, err = fmt.Fprintf(
 			tw,
-			"%s\t%s\t%s\t%s\t%s\n",
+			"%s\t%s\t%s\t%s\t%s\t%s\n",
 			stringid.TruncateID(ctr.Container.ID),
 			ctr.Container.Config.Image,
 			created,
 			state,
+			ipStr,
 			machine,
 		)
 		if err != nil {

cmd/uncloud/service/logs.go

@@ -0,0 +1,336 @@
+package service
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"os"
+	"slices"
+	"strconv"
+	"strings"
+	"time"
+
+	"github.com/charmbracelet/lipgloss"
+	mapset "github.com/deckarep/golang-set/v2"
+	"github.com/docker/docker/pkg/stringid"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/api"
+	"github.com/psviderski/uncloud/pkg/client"
+	"github.com/psviderski/uncloud/pkg/client/compose"
+	"github.com/spf13/cobra"
+)
+
+type logsOptions struct {
+	files    []string
+	follow   bool
+	tail     string
+	since    string
+	until    string
+	utc      bool
+	machines []string
+}
+
+func NewLogsCommand(groupID string) *cobra.Command {
+	var options logsOptions
+
+	cmd := &cobra.Command{
+		Use:     "logs [SERVICE...]",
+		Aliases: []string{"log"},
+		Short:   "View service logs.",
+		Long: `View logs from all replicas of the specified service(s) across all machines in the cluster.
+
+If no services are specified, streams logs from all services defined in the Compose file
+(compose.yaml by default or the file(s) specified with --file).`,
+		Example: `  # View recent logs for a service.
+  uc logs web
+
+  # Stream logs in real-time (follow mode).
+  uc logs -f web
+
+  # View logs from multiple services.
+  uc logs web api db
+
+  # View logs from all services in compose.yaml.
+  uc logs
+
+  # Show last 20 lines per replica (default is 100).
+  uc logs -n 20 web
+
+  # Show all logs without line limit.
+  uc logs -n all web
+
+  # View logs from a specific time range.
+  uc logs --since 3h --until 1h30m web
+
+  # View logs only from replicas running on specific machines.
+  uc logs -m machine1,machine2 web api`,
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			return runLogs(cmd.Context(), uncli, args, options)
+		},
+		GroupID: groupID,
+	}
+
+	cmd.Flags().StringSliceVar(&options.files, "file", nil,
+		"One or more Compose files to load service names from when no services are specified. (default compose.yaml)")
+	cmd.Flags().BoolVarP(&options.follow, "follow", "f", false,
+		"Continually stream new logs.")
+	cmd.Flags().StringSliceVarP(&options.machines, "machine", "m", nil,
+		"Filter logs by machine name or ID. Can be specified multiple times or as a comma-separated list.")
+	cmd.Flags().StringVar(&options.since, "since", "",
+		"Show logs generated on or after the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.\n"+
+			"Examples:\n"+
+			"  --since 2m30s                      Relative duration (2 minutes 30 seconds ago)\n"+
+			"  --since 1h                         Relative duration (1 hour ago)\n"+
+			"  --since 2025-11-24                 RFC 3339 date only (midnight using local timezone)\n"+
+			"  --since 2024-05-14T22:50:00        RFC 3339 date/time using local timezone\n"+
+			"  --since 2024-01-31T10:30:00Z       RFC 3339 date/time in UTC\n"+
+			"  --since 1763953966                 Unix timestamp (seconds since January 1, 1970)")
+	cmd.Flags().StringVarP(&options.tail, "tail", "n", "100",
+		"Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs.")
+	cmd.Flags().StringVar(&options.until, "until", "",
+		"Show logs generated before the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.\n"+
+			"See --since for examples.")
+	cmd.Flags().BoolVar(&options.utc, "utc", false,
+		"Print timestamps in UTC instead of local timezone.")
+
+	return cmd
+}
+
+func runLogs(ctx context.Context, uncli *cli.CLI, serviceNames []string, opts logsOptions) error {
+	// If no services specified, try to load them from the Compose file(s).
+	if len(serviceNames) == 0 {
+		project, err := compose.LoadProject(ctx, opts.files)
+		if err != nil {
+			return fmt.Errorf("load compose file(s): %w", err)
+		}
+		// View logs for all services, including disabled by inactive profiles.
+		serviceNames = append(project.ServiceNames(), project.DisabledServiceNames()...)
+		if len(serviceNames) == 0 {
+			return errors.New("no services found in compose file(s)")
+		}
+	}
+
+	// Parse tail option.
+	tail := -1
+	if opts.tail != "all" {
+		tailInt, err := strconv.Atoi(opts.tail)
+		if err != nil {
+			return fmt.Errorf("invalid --tail value '%s': %w", opts.tail, err)
+		}
+		tail = tailInt
+	}
+
+	c, err := uncli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer c.Close()
+
+	logsOpts := api.ServiceLogsOptions{
+		Follow:   opts.follow,
+		Tail:     tail,
+		Since:    opts.since,
+		Until:    opts.until,
+		Machines: cli.ExpandCommaSeparatedValues(opts.machines),
+	}
+
+	// Collect log streams from all services.
+	machineIDsSet := mapset.NewSet[string]()
+	svcStreams := make([]<-chan api.ServiceLogEntry, 0, len(serviceNames))
+	for _, serviceName := range serviceNames {
+		svc, ch, err := c.ServiceLogs(ctx, serviceName, logsOpts)
+		if err != nil {
+			return fmt.Errorf("stream logs for service '%s': %w", serviceName, err)
+		}
+		svcStreams = append(svcStreams, ch)
+
+		machineIDs := svc.MachineIDs()
+		machineIDsSet.Append(machineIDs...)
+	}
+
+	var stream <-chan api.ServiceLogEntry
+	if len(serviceNames) == 1 {
+		stream = svcStreams[0]
+	} else {
+		// Merge all service streams into a single sorted stream without stall detection as its handled per-service.
+		merger := client.NewLogMerger(svcStreams, client.LogMergerOptions{})
+		stream = merger.Stream()
+	}
+
+	// Fetch machine names for all machines (machineIDsSet) service containers are running on.
+	machines, err := c.ListMachines(ctx, &api.MachineFilter{NamesOrIDs: machineIDsSet.ToSlice()})
+	if err != nil {
+		return fmt.Errorf("list machines: %w", err)
+	}
+	machineNames := make([]string, 0, len(machines))
+	for _, m := range machines {
+		machineNames = append(machineNames, m.Machine.Name)
+	}
+
+	formatter := newLogFormatter(machineNames, serviceNames, opts.utc)
+
+	// Print merged logs.
+	for entry := range stream {
+		if entry.Err != nil {
+			formatter.printError(entry)
+			continue
+		}
+		formatter.printEntry(entry)
+	}
+
+	return nil
+}
+
+// Available colors for machine/service differentiation.
+var colorPalette = []lipgloss.Color{
+	lipgloss.Color("10"), // Bright green
+	lipgloss.Color("11"), // Bright yellow
+	lipgloss.Color("12"), // Bright blue
+	lipgloss.Color("13"), // Bright magenta
+	lipgloss.Color("14"), // Bright cyan
+	lipgloss.Color("2"),  // Green
+	lipgloss.Color("3"),  // Yellow
+	lipgloss.Color("4"),  // Blue
+	lipgloss.Color("5"),  // Magenta
+	lipgloss.Color("6"),  // Cyan
+}
+
+// logFormatter handles formatting and printing of log entries with dynamic column alignment.
+type logFormatter struct {
+	machineNames []string
+	serviceNames []string
+
+	maxMachineWidth int
+	maxServiceWidth int
+
+	utc bool
+}
+
+func newLogFormatter(machineNames, serviceNames []string, utc bool) *logFormatter {
+	slices.Sort(machineNames)
+	slices.Sort(serviceNames)
+
+	maxMachineWidth := 0
+	for _, name := range machineNames {
+		if len(name) > maxMachineWidth {
+			maxMachineWidth = len(name)
+		}
+	}
+
+	maxServiceWidth := 0
+	for _, name := range serviceNames {
+		if len(name) > maxServiceWidth {
+			maxServiceWidth = len(name)
+		}
+	}
+
+	return &logFormatter{
+		machineNames:    machineNames,
+		serviceNames:    serviceNames,
+		maxMachineWidth: maxMachineWidth,
+		maxServiceWidth: maxServiceWidth,
+		utc:             utc,
+	}
+}
+
+// formatTimestamp formats timestamp using local timezone or UTC if configured.
+func (f *logFormatter) formatTimestamp(t time.Time) string {
+	if f.utc {
+		t = t.UTC()
+	} else {
+		t = t.In(time.Local)
+	}
+	dimStyle := lipgloss.NewStyle().Faint(true)
+
+	return dimStyle.Render(t.Format(time.StampMilli))
+}
+
+func (f *logFormatter) formatMachine(name string) string {
+	style := lipgloss.NewStyle().Bold(true).PaddingRight(f.maxMachineWidth - len(name))
+
+	if len(f.serviceNames) == 1 {
+		// Machine name is coloured for single-service logs.
+		i := slices.Index(f.machineNames, name)
+		if i == -1 {
+			f.machineNames = append(f.machineNames, name)
+			i = len(f.machineNames) - 1
+		}
+
+		style = style.Foreground(colorPalette[i%len(colorPalette)])
+	}
+
+	return style.Render(name)
+}
+
+func (f *logFormatter) formatServiceContainer(serviceName, containerID string) string {
+	styleService := lipgloss.NewStyle().Bold(true).PaddingRight(f.maxServiceWidth - len(serviceName))
+	styleContainer := lipgloss.NewStyle().Faint(true)
+
+	if len(f.serviceNames) > 1 {
+		// Service name is coloured for multi-service logs.
+		i := slices.Index(f.serviceNames, serviceName)
+		if i == -1 {
+			f.serviceNames = append(f.serviceNames, serviceName)
+			i = len(f.serviceNames) - 1
+		}
+
+		styleService = styleService.Foreground(colorPalette[i%len(colorPalette)])
+	}
+
+	return styleService.Render(serviceName) + styleContainer.Render("["+containerID[:5]+"]")
+}
+
+// printEntry prints a single log entry with proper formatting.
+func (f *logFormatter) printEntry(entry api.ServiceLogEntry) {
+	if entry.Stream != api.LogStreamStdout && entry.Stream != api.LogStreamStderr {
+		return
+	}
+
+	var output strings.Builder
+
+	// Timestamp
+	output.WriteString(f.formatTimestamp(entry.Timestamp))
+	output.WriteString(" ")
+
+	// Machine name
+	output.WriteString(f.formatMachine(entry.Metadata.MachineName))
+	output.WriteString(" ")
+
+	// Service[container_id]
+	output.WriteString(f.formatServiceContainer(entry.Metadata.ServiceName, entry.Metadata.ContainerID))
+	output.WriteString(" ")
+
+	// Message
+	output.Write(entry.Message)
+
+	// Print to appropriate stream.
+	if entry.Stream == api.LogStreamStderr {
+		fmt.Fprint(os.Stderr, output.String())
+	} else {
+		fmt.Print(output.String())
+	}
+}
+
+// printError prints an error entry (e.g., stalled stream warning).
+func (f *logFormatter) printError(entry api.ServiceLogEntry) {
+	if entry.Metadata.ContainerID != "" {
+		msg := fmt.Sprintf("WARNING: log stream from %s[%s] on machine '%s'",
+			entry.Metadata.ServiceName,
+			stringid.TruncateID(entry.Metadata.ContainerID),
+			entry.Metadata.MachineName)
+
+		if errors.Is(entry.Err, api.ErrLogStreamStalled) {
+			msg += " stopped responding"
+		} else {
+			msg += fmt.Sprintf(": %v", entry.Err)
+		}
+
+		style := lipgloss.NewStyle().Bold(true).Foreground(lipgloss.Color("11")) // Bold bright yellow
+		fmt.Fprintln(os.Stderr, style.Render(msg))
+	} else {
+		msg := fmt.Sprintf("ERROR: %v", entry.Err)
+		style := lipgloss.NewStyle().Bold(true).Foreground(lipgloss.Color("9")) // Bold bright red
+		fmt.Fprintln(os.Stderr, style.Render(msg))
+	}
+}

cmd/uncloud/service/ls.go

@@ -4,34 +4,31 @@ import (
 	"context"
 	"fmt"
 	"os"
+	"slices"
 	"strings"
 	"text/tabwriter"
 
 	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/api"
 	"github.com/spf13/cobra"
 )
 
-func NewListCommand() *cobra.Command {
-	// TODO(lhf): rename to context
-	var cluster string
+func NewListCommand(groupID string) *cobra.Command {
 	cmd := &cobra.Command{
 		Use:     "ls",
 		Aliases: []string{"list"},
 		Short:   "List services.",
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-			return list(cmd.Context(), uncli, cluster)
+			return list(cmd.Context(), uncli)
 		},
+		GroupID: groupID,
 	}
-	cmd.Flags().StringVarP(
-		&cluster, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 	return cmd
 }
 
-func list(ctx context.Context, uncli *cli.CLI, clusterName string) error {
-	client, err := uncli.ConnectCluster(ctx, clusterName)
+func list(ctx context.Context, uncli *cli.CLI) error {
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
@@ -42,15 +39,15 @@ func list(ctx context.Context, uncli *cli.CLI, clusterName string) error {
 		return fmt.Errorf("list services: %w", err)
 	}
 
-	serviceNames := make(map[string]struct{}, len(services))
+	// Sort services by name.
 	haveDuplicateNames := false
-	for _, svc := range services {
-		if _, exists := serviceNames[svc.Name]; exists {
+	slices.SortFunc(services, func(a, b api.Service) int {
+		if a.Name == b.Name {
 			haveDuplicateNames = true
-			break
-		}
-		serviceNames[svc.Name] = struct{}{}
+			return strings.Compare(a.ID, b.ID)
 		}
+		return strings.Compare(a.Name, b.Name)
+	})
 
 	// Print the list of services in a table format.
 	tw := tabwriter.NewWriter(os.Stdout, 0, 0, 3, ' ', 0)
@@ -61,19 +58,29 @@ func list(ctx context.Context, uncli *cli.CLI, clusterName string) error {
 			return fmt.Errorf("write header: %w", err)
 		}
 	}
-	if _, err = fmt.Fprintln(tw, "NAME\tMODE\tREPLICAS\tENDPOINTS"); err != nil {
+	if _, err = fmt.Fprintln(tw, "NAME\tMODE\tREPLICAS\tIMAGE\tENDPOINTS"); err != nil {
 		return fmt.Errorf("write header: %w", err)
 	}
 	for _, s := range services {
-		endpointsSlice := s.Endpoints()
-		endpoints := strings.Join(endpointsSlice, ", ")
+		images := strings.Join(s.Images(), ", ")
+		endpoints := strings.Join(s.Endpoints(), ", ")
+
+		// If no endpoints from ports, check if the service uses custom Caddy config.
+		if endpoints == "" {
+			for _, ctr := range s.Containers {
+				if ctr.Container.ServiceSpec.CaddyConfig() != "" {
+					endpoints = "(custom Caddy config)"
+				}
+			}
+		}
 
 		if haveDuplicateNames {
 			if _, err = fmt.Fprintf(tw, "%s\t", s.ID); err != nil {
 				return fmt.Errorf("write row: %w", err)
 			}
 		}
-		if _, err = fmt.Fprintf(tw, "%s\t%s\t%d\t%s\n", s.Name, s.Mode, len(s.Containers), endpoints); err != nil {
+		if _, err = fmt.Fprintf(tw, "%s\t%s\t%d\t%s\t%s\n",
+			s.Name, s.Mode, len(s.Containers), images, endpoints); err != nil {
 			return fmt.Errorf("write row: %w", err)
 		}
 	}

cmd/uncloud/service/rm.go

@@ -11,31 +11,32 @@ import (
 
 type rmOptions struct {
 	services []string
-	cluster  string
 }
 
-func NewRmCommand() *cobra.Command {
+func NewRmCommand(groupID string) *cobra.Command {
 	opts := rmOptions{}
 	cmd := &cobra.Command{
 		Use:     "rm SERVICE [SERVICE...]",
 		Aliases: []string{"remove", "delete"},
 		Short:   "Remove one or more services.",
+		Long: `Remove one or more services.
+
+The volumes used by the services are preserved and should be removed separately
+with 'uc volume rm'. Anonymous Docker volumes (automatically created from VOLUME
+directives in image Dockerfiles) are automatically removed with their containers.`,
 		Args: cobra.MinimumNArgs(1),
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
 			opts.services = args
 			return rm(cmd.Context(), uncli, opts)
 		},
+		GroupID: groupID,
 	}
-	cmd.Flags().StringVarP(
-		&opts.cluster, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 	return cmd
 }
 
 func rm(ctx context.Context, uncli *cli.CLI, opts rmOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.cluster)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
@@ -50,5 +51,5 @@ func rm(ctx context.Context, uncli *cli.CLI, opts rmOptions) error {
 		}, uncli.ProgressOut(), "Removing service "+s)
 	}
 
-	return nil
+	return err
 }

cmd/uncloud/service/root.go

@@ -8,14 +8,18 @@ func NewRootCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:     "service",
 		Aliases: []string{"svc"},
-		Short:   "Manage services in an Uncloud cluster.",
+		Short:   "Manage services in the cluster.",
 	}
 	cmd.AddCommand(
-		NewInspectCommand(),
-		NewListCommand(),
-		NewRmCommand(),
-		NewRunCommand(),
-		NewScaleCommand(),
+		NewExecCommand(""),
+		NewInspectCommand(""),
+		NewListCommand(""),
+		NewLogsCommand(""),
+		NewRmCommand(""),
+		NewRunCommand(""),
+		NewScaleCommand(""),
+		NewStartCommand(""),
+		NewStopCommand(""),
 	)
 	return cmd
 }

cmd/uncloud/service/run.go

@@ -34,11 +34,9 @@ type runOptions struct {
 	replicas          uint
 	user              string
 	volumes           []string
-
-	cluster string
 }
 
-func NewRunCommand() *cobra.Command {
+func NewRunCommand(groupID string) *cobra.Command {
 	opts := runOptions{}
 
 	cmd := &cobra.Command{
@@ -56,6 +54,7 @@ func NewRunCommand() *cobra.Command {
 
 			return run(cmd.Context(), uncli, opts)
 		},
+		GroupID: groupID,
 	}
 
 	cmd.Flags().StringVar(&opts.caddyfile, "caddyfile", "",
@@ -112,11 +111,6 @@ func NewRunCommand() *cobra.Command {
 			"  -v /data/uploads:/app/uploads         	 Bind mount /data/uploads host directory to /app/uploads in container\n"+
 			"  -v /host/path:/container/path:ro 		 Bind mount a host directory or file as read-only")
 
-	cmd.Flags().StringVarP(
-		&opts.cluster, "context", "c", "",
-		"Name of the cluster context to run the service in. (default is the current context)",
-	)
-
 	return cmd
 }
 
@@ -126,7 +120,7 @@ func run(ctx context.Context, uncli *cli.CLI, opts runOptions) error {
 		return err
 	}
 
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.cluster)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/service/scale.go

@@ -15,10 +15,9 @@ import (
 type scaleOptions struct {
 	service  string
 	replicas uint
-	cluster  string
 }
 
-func NewScaleCommand() *cobra.Command {
+func NewScaleCommand(groupID string) *cobra.Command {
 	opts := scaleOptions{}
 	cmd := &cobra.Command{
 		Use:   "scale SERVICE REPLICAS",
@@ -37,13 +36,9 @@ func NewScaleCommand() *cobra.Command {
 
 			return scale(cmd.Context(), uncli, opts)
 		},
+		GroupID: groupID,
 	}
 
-	cmd.Flags().StringVarP(
-		&opts.cluster, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
-
 	return cmd
 }
 
@@ -57,7 +52,7 @@ func scale(ctx context.Context, uncli *cli.CLI, opts scaleOptions) error {
 		)
 	}
 
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.cluster)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/service/start.go

@@ -0,0 +1,54 @@
+//nolint:dupl
+package service
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/docker/compose/v2/pkg/progress"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/spf13/cobra"
+)
+
+type startOptions struct {
+	services []string
+}
+
+func NewStartCommand(groupID string) *cobra.Command {
+	opts := startOptions{}
+	cmd := &cobra.Command{
+		Use:   "start SERVICE [SERVICE...]",
+		Short: "Start one or more services.",
+		Long: `Start one or more previously stopped services.
+
+Starts all containers of the specified service(s) across all machines in the cluster.
+Services can be specified by name or ID.`,
+		Args: cobra.MinimumNArgs(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			opts.services = args
+			return start(cmd.Context(), uncli, opts)
+		},
+		GroupID: groupID,
+	}
+	return cmd
+}
+
+func start(ctx context.Context, uncli *cli.CLI, opts startOptions) error {
+	client, err := uncli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer client.Close()
+
+	for _, s := range opts.services {
+		err = progress.RunWithTitle(ctx, func(ctx context.Context) error {
+			if err = client.StartService(ctx, s); err != nil {
+				return fmt.Errorf("start service '%s': %w", s, err)
+			}
+			return nil
+		}, uncli.ProgressOut(), "Starting service "+s)
+	}
+
+	return err
+}

cmd/uncloud/service/stop.go

@@ -0,0 +1,68 @@
+//nolint:dupl
+package service
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/docker/compose/v2/pkg/progress"
+	"github.com/docker/docker/api/types/container"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/spf13/cobra"
+)
+
+type stopOptions struct {
+	services []string
+	signal   string
+	timeout  int
+}
+
+func NewStopCommand(groupID string) *cobra.Command {
+	opts := stopOptions{}
+	cmd := &cobra.Command{
+		Use:   "stop SERVICE [SERVICE...]",
+		Short: "Stop one or more services.",
+		Long: `Stop one or more running services.
+
+Gracefully stops all containers of the specified service(s) across all machines in the cluster.
+Services can be specified by name or ID. Stopped services can be restarted with 'uc start'.`,
+		Args: cobra.MinimumNArgs(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			opts.services = args
+			return stop(cmd.Context(), uncli, opts)
+		},
+		GroupID: groupID,
+	}
+	cmd.Flags().StringVarP(&opts.signal, "signal", "s", "",
+		"Signal to send to each container's main process.\n"+
+			"Can be a signal name (SIGTERM, SIGINT, SIGHUP, etc.) or a number. (default SIGTERM)")
+	cmd.Flags().IntVarP(&opts.timeout, "timeout", "t", 10,
+		"Seconds to wait for each container to stop gracefully before forcibly killing it with SIGKILL.\n"+
+			"Use -1 to wait indefinitely.")
+	return cmd
+}
+
+func stop(ctx context.Context, uncli *cli.CLI, opts stopOptions) error {
+	client, err := uncli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer client.Close()
+
+	stopOpts := container.StopOptions{
+		Signal:  opts.signal,
+		Timeout: &opts.timeout,
+	}
+
+	for _, s := range opts.services {
+		err = progress.RunWithTitle(ctx, func(ctx context.Context) error {
+			if err = client.StopService(ctx, s, stopOpts); err != nil {
+				return fmt.Errorf("stop service '%s': %w", s, err)
+			}
+			return nil
+		}, uncli.ProgressOut(), "Stopping service "+s)
+	}
+
+	return err
+}

cmd/uncloud/volume/create.go

@@ -19,7 +19,6 @@ type createOptions struct {
 	driverOpts []string
 	labels     []string
 	machine    string
-	context    string
 }
 
 func NewCreateCommand() *cobra.Command {
@@ -50,14 +49,12 @@ func NewCreateCommand() *cobra.Command {
 		"Labels to assign to the volume in the form of 'key=value' pairs. Can be specified multiple times.")
 	cmd.Flags().StringVarP(&opts.machine, "machine", "m", "",
 		"Name or ID of the machine to create the volume on.")
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
 
 	return cmd
 }
 
 func create(ctx context.Context, uncli *cli.CLI, name string, opts createOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/volume/inspect.go

@@ -13,7 +13,6 @@ import (
 
 type inspectOptions struct {
 	machine string
-	context string
 }
 
 func NewInspectCommand() *cobra.Command {
@@ -32,14 +31,12 @@ func NewInspectCommand() *cobra.Command {
 	cmd.Flags().StringVarP(&opts.machine, "machine", "m", "",
 		"Name or ID of the machine where the volume is located. "+
 			"If not specified, the volume will be searched across all machines.")
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
 
 	return cmd
 }
 
 func inspect(ctx context.Context, uncli *cli.CLI, name string, opts inspectOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/volume/ls.go

@@ -16,7 +16,6 @@ import (
 type listOptions struct {
 	machines []string
 	quiet    bool
-	context  string
 }
 
 func NewListCommand() *cobra.Command {
@@ -38,14 +37,11 @@ func NewListCommand() *cobra.Command {
 	cmd.Flags().BoolVarP(&opts.quiet, "quiet", "q", false,
 		"Only display volume names.")
 
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
-
 	return cmd
 }
 
 func list(ctx context.Context, uncli *cli.CLI, opts listOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/volume/rm.go

@@ -14,7 +14,6 @@ type removeOptions struct {
 	force    bool
 	machines []string
 	yes      bool
-	context  string
 }
 
 func NewRemoveCommand() *cobra.Command {
@@ -41,14 +40,11 @@ func NewRemoveCommand() *cobra.Command {
 	cmd.Flags().BoolVarP(&opts.yes, "yes", "y", false,
 		"Do not prompt for confirmation before removing the volume(s).")
 
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
-
 	return cmd
 }
 
 func remove(ctx context.Context, uncli *cli.CLI, names []string, opts removeOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/volume/root.go

@@ -7,7 +7,7 @@ import (
 func NewRootCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:   "volume",
-		Short: "Manage volumes in an Uncloud cluster.",
+		Short: "Manage volumes in the cluster.",
 	}
 	cmd.AddCommand(
 		NewCreateCommand(),

cmd/uncloudd/dialstdio.go

@@ -0,0 +1,128 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"net"
+	"os"
+
+	"github.com/psviderski/uncloud/internal/machine"
+	"github.com/spf13/cobra"
+)
+
+func newDialStdioCommand() *cobra.Command {
+	var socketPath string
+
+	cmd := &cobra.Command{
+		Use:    "dial-stdio",
+		Short:  "Proxy stdin/stdout to the Uncloud API socket",
+		Hidden: true,
+		RunE: func(cmd *cobra.Command, args []string) error {
+			return runDialStdio(cmd.Context(), socketPath, os.Stdin, os.Stdout)
+		},
+	}
+
+	cmd.Flags().StringVar(&socketPath, "socket", machine.DefaultUncloudSockPath,
+		"Path to the Uncloud API socket")
+
+	return cmd
+}
+
+// halfReadCloser is the read side of a half-duplex connection.
+type halfReadCloser interface {
+	io.Reader
+	CloseRead() error
+}
+
+// halfWriteCloser is the write side of a half-duplex connection.
+type halfWriteCloser interface {
+	io.Writer
+	CloseWrite() error
+}
+
+// halfReadCloserWrapper wraps an io.ReadCloser to implement halfReadCloser.
+type halfReadCloserWrapper struct {
+	io.ReadCloser
+}
+
+func (x *halfReadCloserWrapper) CloseRead() error {
+	return x.Close()
+}
+
+// halfWriteCloserWrapper wraps an io.WriteCloser to implement halfWriteCloser.
+type halfWriteCloserWrapper struct {
+	io.WriteCloser
+}
+
+func (x *halfWriteCloserWrapper) CloseWrite() error {
+	return x.Close()
+}
+
+func runDialStdio(ctx context.Context, socketPath string, stdin io.Reader, stdout io.Writer) error {
+	// Connect to the unix socket.
+	var dialer net.Dialer
+	conn, err := dialer.DialContext(ctx, "unix", socketPath)
+	if err != nil {
+		return fmt.Errorf("connect to socket %q: %w", socketPath, err)
+	}
+	defer conn.Close()
+
+	// Wrap stdin/stdout to support half-closing.
+	var stdinCloser halfReadCloser
+	if c, ok := stdin.(halfReadCloser); ok {
+		stdinCloser = c
+	} else if c, ok := stdin.(io.ReadCloser); ok {
+		stdinCloser = &halfReadCloserWrapper{c}
+	}
+
+	var stdoutCloser halfWriteCloser
+	if c, ok := stdout.(halfWriteCloser); ok {
+		stdoutCloser = c
+	} else if c, ok := stdout.(io.WriteCloser); ok {
+		stdoutCloser = &halfWriteCloserWrapper{c}
+	}
+
+	// Copy data bidirectionally between stdin/stdout and the socket.
+	stdin2socket := make(chan error, 1)
+	socket2stdout := make(chan error, 1)
+
+	// Copy from stdin to socket.
+	go func() {
+		_, err := io.Copy(conn, stdin)
+		stdin2socket <- err
+		// Close write side of connection after stdin is done.
+		if unixConn, ok := conn.(*net.UnixConn); ok {
+			unixConn.CloseWrite()
+		}
+		if stdinCloser != nil {
+			stdinCloser.CloseRead()
+		}
+	}()
+
+	// Copy from socket to stdout.
+	go func() {
+		_, err := io.Copy(stdout, conn)
+		socket2stdout <- err
+		// Close read side of connection after socket is done sending.
+		if unixConn, ok := conn.(*net.UnixConn); ok {
+			unixConn.CloseRead()
+		}
+		if stdoutCloser != nil {
+			stdoutCloser.CloseWrite()
+		}
+	}()
+
+	select {
+	case err = <-stdin2socket:
+		if err != nil {
+			return err
+		}
+		// wait for stdout
+		err = <-socket2stdout
+	case err = <-socket2stdout:
+		// return immediately, matching Docker's approach
+		// (stdin is never closed when TTY)
+	}
+	return err
+}

cmd/uncloudd/main.go

@@ -42,6 +42,9 @@ func main() {
 		"Directory for storing persistent machine state")
 	_ = cmd.MarkFlagDirname("data-dir")
 
+	// Add dial-stdio subcommand.
+	cmd.AddCommand(newDialStdioCommand())
+
 	// ctx is canceled when the daemon command is interrupted.
 	ctx, cancel := context.WithCancel(context.Background())
 

go.mod

@@ -1,25 +1,30 @@
 module github.com/psviderski/uncloud
 
-go 1.23.0
+go 1.25
 
 require (
-	github.com/BurntSushi/toml v1.4.0
+	github.com/BurntSushi/toml v1.5.0
 	github.com/Masterminds/semver v1.5.0
 	github.com/Masterminds/squirrel v1.5.4
 	github.com/alecthomas/chroma/v2 v2.20.0
 	github.com/caddyserver/caddy/v2 v2.8.4
 	github.com/cenkalti/backoff/v4 v4.3.0
+	github.com/charmbracelet/bubbles v0.21.0
+	github.com/charmbracelet/bubbletea v1.3.10
 	github.com/charmbracelet/huh v0.6.0
-	github.com/charmbracelet/lipgloss v0.13.0
-	github.com/compose-spec/compose-go/v2 v2.4.5
+	github.com/charmbracelet/huh/spinner v0.0.0-20251110114415-25888d17260b
+	github.com/charmbracelet/lipgloss v1.1.0
+	github.com/compose-spec/compose-go/v2 v2.9.0
+	github.com/containerd/errdefs v1.0.0
+	github.com/containerd/platforms v1.0.0-rc.1
 	github.com/coreos/go-systemd v0.0.0-20191104093116-d3cd4ed1dbcf
 	github.com/deckarep/golang-set/v2 v2.8.0
 	github.com/dgraph-io/badger/v3 v3.2103.5
 	github.com/distribution/reference v0.6.0
-	github.com/docker/cli v27.5.0+incompatible
-	github.com/docker/compose/v2 v2.31.0
-	github.com/docker/docker v27.4.0-rc.2+incompatible
-	github.com/docker/go-connections v0.5.0
+	github.com/docker/cli v28.5.0+incompatible
+	github.com/docker/compose/v2 v2.40.0
+	github.com/docker/docker v28.5.0+incompatible
+	github.com/docker/go-connections v0.6.0
 	github.com/docker/go-units v0.5.0
 	github.com/goccy/go-yaml v1.17.1
 	github.com/google/go-cmp v0.7.0
@@ -37,45 +42,66 @@ require (
 	github.com/lmittmann/tint v1.0.5
 	github.com/miekg/dns v1.1.65
 	github.com/mitchellh/mapstructure v1.5.0
-	github.com/moby/term v0.5.0
+	github.com/moby/term v0.5.2
+	github.com/muesli/termenv v0.16.0
 	github.com/opencontainers/go-digest v1.0.0
-	github.com/opencontainers/image-spec v1.1.0
+	github.com/opencontainers/image-spec v1.1.1
+	github.com/psviderski/unregistry v0.4.1
 	github.com/siderolabs/discovery-api v0.1.4
 	github.com/siderolabs/discovery-client v0.1.9
 	github.com/siderolabs/grpc-proxy v0.5.1
-	github.com/spf13/cobra v1.8.1
-	github.com/stretchr/testify v1.10.0
-	github.com/vishvananda/netlink v1.3.0
+	github.com/spf13/cobra v1.10.1
+	github.com/stretchr/testify v1.11.1
+	github.com/vishvananda/netlink v1.3.1
 	go.uber.org/zap v1.27.0
 	go4.org/netipx v0.0.0-20231129151722-fdeea329fbba
-	golang.org/x/crypto v0.33.0
-	golang.org/x/net v0.35.0
-	golang.org/x/sync v0.11.0
-	golang.org/x/sys v0.31.0
+	golang.org/x/crypto v0.41.0
+	golang.org/x/net v0.43.0
+	golang.org/x/sync v0.17.0
+	golang.org/x/sys v0.36.0
+	golang.org/x/term v0.34.0
 	golang.zx2c4.com/wireguard v0.0.0-20231211153847-12269c276173
 	golang.zx2c4.com/wireguard/wgctrl v0.0.0-20230429144221-925a1e7659e6
-	google.golang.org/genproto/googleapis/rpc v0.0.0-20241209162323-e6fa225c2576
-	google.golang.org/grpc v1.68.1
-	google.golang.org/protobuf v1.36.3
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20250528174236-200df99c418a
+	google.golang.org/grpc v1.74.2
+	google.golang.org/protobuf v1.36.9
 	modernc.org/sqlite v1.36.3
 )
 
 require (
+	cloud.google.com/go/kms v1.19.0 // indirect
+	cloud.google.com/go/longrunning v0.6.0 // indirect
 	filippo.io/edwards25519 v1.1.0 // indirect
-	github.com/AdaLogics/go-fuzz-headers v0.0.0-20230811130428-ced1acdcaa24 // indirect
+	github.com/AlecAivazis/survey/v2 v2.3.7 // indirect
 	github.com/AndreasBriese/bbloom v0.0.0-20190825152654-46b345b51c96 // indirect
-	github.com/Azure/go-ansiterm v0.0.0-20210617225240-d185dfc1b5a1 // indirect
+	github.com/Azure/go-ansiterm v0.0.0-20250102033503-faa5f7b0171c // indirect
+	github.com/DefangLabs/secret-detector v0.0.0-20250403165618-22662109213e // indirect
 	github.com/Masterminds/goutils v1.1.1 // indirect
-	github.com/Masterminds/semver/v3 v3.2.1 // indirect
+	github.com/Masterminds/semver/v3 v3.4.0 // indirect
 	github.com/Masterminds/sprig/v3 v3.2.3 // indirect
 	github.com/Microsoft/go-winio v0.6.2 // indirect
-	github.com/OneOfOne/xxhash v1.2.8 // indirect
+	github.com/Microsoft/hcsshim v0.13.0 // indirect
+	github.com/acarl005/stripansi v0.0.0-20180116102854-5a71ef0e047d // indirect
 	github.com/antlr4-go/antlr/v4 v4.13.0 // indirect
+	github.com/apparentlymart/go-textseg/v15 v15.0.0 // indirect
 	github.com/armon/circbuf v0.0.0-20190214190532-5111143e8da2 // indirect
 	github.com/armon/go-metrics v0.4.1 // indirect
 	github.com/armon/go-radix v1.0.0 // indirect
 	github.com/aryann/difflib v0.0.0-20210328193216-ff5ff6dc229b // indirect
 	github.com/atotto/clipboard v0.1.4 // indirect
+	github.com/aws/aws-sdk-go-v2 v1.30.3 // indirect
+	github.com/aws/aws-sdk-go-v2/config v1.27.27 // indirect
+	github.com/aws/aws-sdk-go-v2/credentials v1.17.27 // indirect
+	github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.16.11 // indirect
+	github.com/aws/aws-sdk-go-v2/internal/configsources v1.3.15 // indirect
+	github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.6.15 // indirect
+	github.com/aws/aws-sdk-go-v2/internal/ini v1.8.0 // indirect
+	github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.11.3 // indirect
+	github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.11.17 // indirect
+	github.com/aws/aws-sdk-go-v2/service/sso v1.22.4 // indirect
+	github.com/aws/aws-sdk-go-v2/service/ssooidc v1.26.4 // indirect
+	github.com/aws/aws-sdk-go-v2/service/sts v1.30.3 // indirect
+	github.com/aws/smithy-go v1.20.3 // indirect
 	github.com/aymanbagabas/go-osc52/v2 v2.0.1 // indirect
 	github.com/beorn7/perks v1.0.1 // indirect
 	github.com/bgentry/speakeasy v0.2.0 // indirect
@@ -85,68 +111,86 @@ require (
 	github.com/catppuccin/go v0.2.0 // indirect
 	github.com/cespare/xxhash v1.1.0 // indirect
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
-	github.com/charmbracelet/bubbles v0.20.0 // indirect
-	github.com/charmbracelet/bubbletea v1.1.1 // indirect
-	github.com/charmbracelet/x/ansi v0.3.2 // indirect
+	github.com/charmbracelet/colorprofile v0.2.3-0.20250311203215-f60798e515dc // indirect
+	github.com/charmbracelet/x/ansi v0.10.1 // indirect
+	github.com/charmbracelet/x/cellbuf v0.0.13-0.20250311204145-2c3ea96c31dd // indirect
 	github.com/charmbracelet/x/exp/strings v0.0.0-20240919170804-a4978c8e603a // indirect
-	github.com/charmbracelet/x/term v0.2.0 // indirect
+	github.com/charmbracelet/x/term v0.2.1 // indirect
 	github.com/chzyer/readline v1.5.1 // indirect
-	github.com/cloudflare/cfssl v1.6.4 // indirect
-	github.com/containerd/console v1.0.4 // indirect
-	github.com/containerd/containerd v1.7.24 // indirect
-	github.com/containerd/containerd/api v1.7.19 // indirect
-	github.com/containerd/continuity v0.4.4 // indirect
-	github.com/containerd/errdefs v0.3.0 // indirect
+	github.com/containerd/cgroups/v3 v3.0.5 // indirect
+	github.com/containerd/console v1.0.5 // indirect
+	github.com/containerd/containerd/api v1.9.0 // indirect
+	github.com/containerd/containerd/v2 v2.1.4 // indirect
+	github.com/containerd/continuity v0.4.5 // indirect
+	github.com/containerd/errdefs/pkg v0.3.0 // indirect
+	github.com/containerd/fifo v1.1.0 // indirect
 	github.com/containerd/log v0.1.0 // indirect
-	github.com/containerd/platforms v0.2.1 // indirect
+	github.com/containerd/plugin v1.0.0 // indirect
 	github.com/containerd/stargz-snapshotter/estargz v0.16.3 // indirect
-	github.com/containerd/ttrpc v1.2.5 // indirect
-	github.com/containerd/typeurl/v2 v2.2.0 // indirect
-	github.com/cpuguy83/go-md2man/v2 v2.0.6 // indirect
-	github.com/davecgh/go-spew v1.1.1 // indirect
+	github.com/containerd/ttrpc v1.2.7 // indirect
+	github.com/containerd/typeurl/v2 v2.2.3 // indirect
+	github.com/cpuguy83/go-md2man/v2 v2.0.7 // indirect
+	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc // indirect
 	github.com/decred/dcrd/dcrec/secp256k1/v4 v4.3.0 // indirect
 	github.com/dgraph-io/badger v1.6.2 // indirect
 	github.com/dgraph-io/badger/v2 v2.2007.4 // indirect
 	github.com/dgraph-io/ristretto v0.1.1 // indirect
 	github.com/dgryski/go-farm v0.0.0-20200201041132-a6ae2369ad13 // indirect
+	github.com/dgryski/go-rendezvous v0.0.0-20200823014737-9f7001d12a5f // indirect
+	github.com/distribution/distribution/v3 v3.0.0 // indirect
 	github.com/dlclark/regexp2 v1.11.5 // indirect
-	github.com/docker/buildx v0.18.0 // indirect
+	github.com/docker/buildx v0.29.1 // indirect
+	github.com/docker/cli-docs-tool v0.10.0 // indirect
 	github.com/docker/distribution v2.8.3+incompatible // indirect
-	github.com/docker/docker-credential-helpers v0.8.2 // indirect
+	github.com/docker/docker-credential-helpers v0.9.3 // indirect
 	github.com/docker/go v1.5.1-1.0.20160303222718-d30aec9fd63c // indirect
 	github.com/docker/go-events v0.0.0-20190806004212-e31b211e4f1c // indirect
 	github.com/docker/go-metrics v0.0.1 // indirect
 	github.com/dustin/go-humanize v1.0.1 // indirect
+	github.com/eiannone/keyboard v0.0.0-20220611211555-0d226195f203 // indirect
+	github.com/emicklei/go-restful/v3 v3.11.0 // indirect
 	github.com/erikgeiser/coninput v0.0.0-20211004153227-1c3628e74d0f // indirect
-	github.com/fatih/color v1.17.0 // indirect
+	github.com/fatih/color v1.18.0 // indirect
 	github.com/felixge/httpsnoop v1.0.4 // indirect
+	github.com/fsnotify/fsevents v0.2.0 // indirect
 	github.com/fvbommel/sortorder v1.1.0 // indirect
+	github.com/fxamacker/cbor/v2 v2.7.0 // indirect
 	github.com/go-jose/go-jose/v3 v3.0.3 // indirect
 	github.com/go-kit/kit v0.13.0 // indirect
 	github.com/go-kit/log v0.2.1 // indirect
 	github.com/go-logfmt/logfmt v0.6.0 // indirect
-	github.com/go-logr/logr v1.4.2 // indirect
+	github.com/go-logr/logr v1.4.3 // indirect
 	github.com/go-logr/stdr v1.2.2 // indirect
+	github.com/go-openapi/jsonpointer v0.21.0 // indirect
+	github.com/go-openapi/jsonreference v0.20.2 // indirect
+	github.com/go-openapi/swag v0.23.0 // indirect
 	github.com/go-sql-driver/mysql v1.8.1 // indirect
 	github.com/go-task/slim-sprig/v3 v3.0.0 // indirect
-	github.com/go-viper/mapstructure/v2 v2.0.0 // indirect
+	github.com/go-viper/mapstructure/v2 v2.4.0 // indirect
 	github.com/godbus/dbus/v5 v5.1.0 // indirect
 	github.com/gofrs/flock v0.12.1 // indirect
 	github.com/gogo/protobuf v1.3.2 // indirect
-	github.com/golang/glog v1.2.2 // indirect
+	github.com/golang-jwt/jwt/v5 v5.2.2 // indirect
+	github.com/golang/glog v1.2.5 // indirect
 	github.com/golang/groupcache v0.0.0-20210331224755-41bb18bfe9da // indirect
 	github.com/golang/protobuf v1.5.4 // indirect
 	github.com/golang/snappy v0.0.4 // indirect
 	github.com/google/btree v1.1.2 // indirect
 	github.com/google/cel-go v0.20.1 // indirect
 	github.com/google/flatbuffers v24.3.25+incompatible // indirect
-	github.com/google/pprof v0.0.0-20250317173921-a4b03ec1a45e // indirect
+	github.com/google/gnostic-models v0.6.8 // indirect
+	github.com/google/gofuzz v1.2.0 // indirect
+	github.com/google/pprof v0.0.0-20250403155104-27863c87afa6 // indirect
 	github.com/google/shlex v0.0.0-20191202100458-e7afc7fbc510 // indirect
 	github.com/google/uuid v1.6.0 // indirect
+	github.com/gorilla/handlers v1.5.2 // indirect
 	github.com/gorilla/mux v1.8.1 // indirect
-	github.com/grpc-ecosystem/grpc-gateway/v2 v2.24.0 // indirect
+	github.com/gorilla/websocket v1.5.3 // indirect
+	github.com/grpc-ecosystem/grpc-gateway/v2 v2.26.1 // indirect
 	github.com/hashicorp/errwrap v1.1.0 // indirect
+	github.com/hashicorp/go-cleanhttp v0.5.2 // indirect
 	github.com/hashicorp/go-immutable-radix v1.3.1 // indirect
+	github.com/hashicorp/go-immutable-radix/v2 v2.1.0 // indirect
 	github.com/hashicorp/go-msgpack v0.5.5 // indirect
 	github.com/hashicorp/go-msgpack/v2 v2.1.2 // indirect
 	github.com/hashicorp/go-multierror v1.1.1 // indirect
@@ -154,13 +198,15 @@ require (
 	github.com/hashicorp/go-syslog v1.0.0 // indirect
 	github.com/hashicorp/go-version v1.7.0 // indirect
 	github.com/hashicorp/golang-lru v1.0.2 // indirect
+	github.com/hashicorp/golang-lru/arc/v2 v2.0.7 // indirect
 	github.com/hashicorp/golang-lru/v2 v2.0.7 // indirect
 	github.com/hashicorp/logutils v1.0.0 // indirect
 	github.com/hashicorp/mdns v1.0.5 // indirect
 	github.com/huandu/xstrings v1.5.0 // indirect
 	github.com/imdario/mergo v0.3.16 // indirect
-	github.com/in-toto/in-toto-golang v0.5.0 // indirect
+	github.com/in-toto/in-toto-golang v0.9.0 // indirect
 	github.com/inconshreveable/mousetrap v1.1.0 // indirect
+	github.com/inhies/go-bytesize v0.0.0-20220417184213-4913239db9cf // indirect
 	github.com/ipfs/bbloom v0.0.4 // indirect
 	github.com/ipfs/go-block-format v0.2.0 // indirect
 	github.com/ipfs/go-ipfs-util v0.0.3 // indirect
@@ -178,8 +224,12 @@ require (
 	github.com/jackc/pgtype v1.14.0 // indirect
 	github.com/jackc/pgx/v4 v4.18.3 // indirect
 	github.com/jbenet/goprocess v0.1.4 // indirect
+	github.com/jonboulle/clockwork v0.5.0 // indirect
+	github.com/josharian/intern v1.0.0 // indirect
 	github.com/josharian/native v1.1.0 // indirect
-	github.com/klauspost/compress v1.17.11 // indirect
+	github.com/json-iterator/go v1.1.12 // indirect
+	github.com/kballard/go-shellquote v0.0.0-20180428030007-95032a82bc51 // indirect
+	github.com/klauspost/compress v1.18.0 // indirect
 	github.com/klauspost/cpuid/v2 v2.2.9 // indirect
 	github.com/lann/builder v0.0.0-20180802200727-47ae307949d0 // indirect
 	github.com/lann/ps v0.0.0-20150810152359-62de8c46ede0 // indirect
@@ -189,6 +239,7 @@ require (
 	github.com/libp2p/go-libp2p-pubsub v0.11.0 // indirect
 	github.com/libp2p/go-msgio v0.3.0 // indirect
 	github.com/lucasb-eyer/go-colorful v1.2.0 // indirect
+	github.com/mailru/easyjson v0.7.7 // indirect
 	github.com/manifoldco/promptui v0.9.0 // indirect
 	github.com/mattn/go-colorable v0.1.13 // indirect
 	github.com/mattn/go-isatty v0.0.20 // indirect
@@ -208,19 +259,26 @@ require (
 	github.com/mitchellh/go-ps v1.0.0 // indirect
 	github.com/mitchellh/hashstructure/v2 v2.0.2 // indirect
 	github.com/mitchellh/reflectwalk v1.0.2 // indirect
-	github.com/moby/buildkit v0.17.2 // indirect
+	github.com/moby/buildkit v0.25.0 // indirect
 	github.com/moby/docker-image-spec v1.3.1 // indirect
+	github.com/moby/go-archive v0.1.0 // indirect
 	github.com/moby/locker v1.0.1 // indirect
 	github.com/moby/patternmatcher v0.6.0 // indirect
+	github.com/moby/spdystream v0.5.0 // indirect
+	github.com/moby/sys/atomicwriter v0.1.0 // indirect
+	github.com/moby/sys/capability v0.4.0 // indirect
+	github.com/moby/sys/mountinfo v0.7.2 // indirect
 	github.com/moby/sys/sequential v0.6.0 // indirect
 	github.com/moby/sys/signal v0.7.1 // indirect
-	github.com/moby/sys/user v0.3.0 // indirect
+	github.com/moby/sys/symlink v0.3.0 // indirect
+	github.com/moby/sys/user v0.4.0 // indirect
 	github.com/moby/sys/userns v0.1.0 // indirect
+	github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd // indirect
+	github.com/modern-go/reflect2 v1.0.2 // indirect
 	github.com/morikuni/aec v1.0.0 // indirect
 	github.com/mr-tron/base58 v1.2.0 // indirect
 	github.com/muesli/ansi v0.0.0-20230316100256-276c6243b2f6 // indirect
 	github.com/muesli/cancelreader v0.2.2 // indirect
-	github.com/muesli/termenv v0.15.3-0.20240618155329-98d742f6907a // indirect
 	github.com/multiformats/go-base32 v0.1.0 // indirect
 	github.com/multiformats/go-base36 v0.2.0 // indirect
 	github.com/multiformats/go-multiaddr v0.13.0 // indirect
@@ -231,94 +289,131 @@ require (
 	github.com/multiformats/go-multistream v0.5.0 // indirect
 	github.com/multiformats/go-varint v0.0.7 // indirect
 	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
+	github.com/mxk/go-flowrate v0.0.0-20140419014527-cca7078d478f // indirect
 	github.com/ncruces/go-strftime v0.1.9 // indirect
 	github.com/onsi/ginkgo/v2 v2.22.0 // indirect
+	github.com/opencontainers/runtime-spec v1.2.1 // indirect
+	github.com/opencontainers/selinux v1.12.0 // indirect
 	github.com/pelletier/go-toml v1.9.5 // indirect
 	github.com/pires/go-proxyproto v0.7.0 // indirect
 	github.com/pkg/errors v0.9.1 // indirect
 	github.com/planetscale/vtprotobuf v0.6.1-0.20240319094008-0393e58bdf10 // indirect
-	github.com/pmezard/go-difflib v1.0.0 // indirect
+	github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2 // indirect
 	github.com/polydawn/refmt v0.89.0 // indirect
 	github.com/posener/complete v1.2.3 // indirect
-	github.com/prometheus/client_golang v1.20.5 // indirect
+	github.com/prometheus/client_golang v1.22.0 // indirect
 	github.com/prometheus/client_model v0.6.1 // indirect
-	github.com/prometheus/common v0.61.0 // indirect
+	github.com/prometheus/common v0.62.0 // indirect
 	github.com/prometheus/procfs v0.15.1 // indirect
 	github.com/quic-go/qpack v0.5.1 // indirect
 	github.com/quic-go/quic-go v0.48.2 // indirect
+	github.com/redis/go-redis/extra/rediscmd/v9 v9.0.5 // indirect
+	github.com/redis/go-redis/extra/redisotel/v9 v9.0.5 // indirect
+	github.com/redis/go-redis/v9 v9.7.3 // indirect
 	github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec // indirect
 	github.com/rivo/uniseg v0.4.7 // indirect
 	github.com/rs/xid v1.5.0 // indirect
 	github.com/russross/blackfriday/v2 v2.1.0 // indirect
+	github.com/santhosh-tekuri/jsonschema/v6 v6.0.1 // indirect
 	github.com/sean-/seed v0.0.0-20170313163322-e2103e2c3529 // indirect
-	github.com/secure-systems-lab/go-securesystemslib v0.4.0 // indirect
+	github.com/secure-systems-lab/go-securesystemslib v0.6.0 // indirect
+	github.com/serialx/hashring v0.0.0-20200727003509-22c0c7ab6b1b // indirect
 	github.com/shibumi/go-pathspec v1.3.0 // indirect
 	github.com/shopspring/decimal v1.4.0 // indirect
 	github.com/shurcooL/sanitized_anchor_name v1.0.0 // indirect
 	github.com/siderolabs/gen v0.4.8 // indirect
 	github.com/sirupsen/logrus v1.9.3 // indirect
+	github.com/skratchdot/open-golang v0.0.0-20200116055534-eef842397966 // indirect
 	github.com/slackhq/nebula v1.6.1 // indirect
 	github.com/smallstep/certificates v0.26.1 // indirect
 	github.com/smallstep/nosql v0.6.1 // indirect
-	github.com/smallstep/pkcs7 v0.0.0-20231024181729-3b98ecc1ca81 // indirect
+	github.com/smallstep/pkcs7 v0.1.1 // indirect
 	github.com/smallstep/scep v0.0.0-20231024192529-aee96d7ad34d // indirect
 	github.com/smallstep/truststore v0.13.0 // indirect
 	github.com/spaolacci/murmur3 v1.1.0 // indirect
 	github.com/spf13/cast v1.7.0 // indirect
-	github.com/spf13/pflag v1.0.5 // indirect
+	github.com/spf13/pflag v1.0.10 // indirect
 	github.com/stoewer/go-strcase v1.2.0 // indirect
 	github.com/stretchr/objx v0.5.2 // indirect
 	github.com/tailscale/tscert v0.0.0-20240517230440-bbccfbf48933 // indirect
 	github.com/theupdateframework/notary v0.7.0 // indirect
-	github.com/tonistiigi/dchapes-mode v0.0.0-20241001053921-ca0759fec205 // indirect
-	github.com/tonistiigi/fsutil v0.0.0-20241028165955-397af5306b5c // indirect
-	github.com/tonistiigi/go-csvvalue v0.0.0-20240710180619-ddb21b71c0b4 // indirect
+	github.com/tilt-dev/fsnotify v1.4.8-0.20220602155310-fff9c274a375 // indirect
+	github.com/tonistiigi/dchapes-mode v0.0.0-20250318174251-73d941a28323 // indirect
+	github.com/tonistiigi/fsutil v0.0.0-20250605211040-586307ad452f // indirect
+	github.com/tonistiigi/go-csvvalue v0.0.0-20240814133006-030d3b2625d0 // indirect
 	github.com/tonistiigi/units v0.0.0-20180711220420-6950e57a87ea // indirect
 	github.com/tonistiigi/vt100 v0.0.0-20240514184818-90bafcd6abab // indirect
-	github.com/urfave/cli v1.22.16 // indirect
-	github.com/vbatts/tar-split v0.11.6 // indirect
-	github.com/vishvananda/netns v0.0.4 // indirect
-	github.com/xeipuuv/gojsonpointer v0.0.0-20190905194746-02993c407bfb // indirect
-	github.com/xeipuuv/gojsonreference v0.0.0-20180127040603-bd5ef7bd5415 // indirect
-	github.com/xeipuuv/gojsonschema v1.2.0 // indirect
+	github.com/urfave/cli v1.22.17 // indirect
+	github.com/vbatts/tar-split v0.12.1 // indirect
+	github.com/vishvananda/netns v0.0.5 // indirect
+	github.com/x448/float16 v0.8.4 // indirect
+	github.com/xhit/go-str2duration/v2 v2.1.0 // indirect
+	github.com/xo/terminfo v0.0.0-20220910002029-abceb7e1c41e // indirect
+	github.com/zclconf/go-cty v1.17.0 // indirect
 	github.com/zeebo/blake3 v0.2.4 // indirect
-	go.etcd.io/bbolt v1.3.11 // indirect
+	go.etcd.io/bbolt v1.4.3 // indirect
 	go.opencensus.io v0.24.0 // indirect
 	go.opentelemetry.io/auto/sdk v1.1.0 // indirect
-	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.49.0 // indirect
-	go.opentelemetry.io/contrib/instrumentation/net/http/httptrace/otelhttptrace v0.46.1 // indirect
-	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.58.0 // indirect
-	go.opentelemetry.io/otel v1.33.0 // indirect
-	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v0.44.0 // indirect
-	go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.33.0 // indirect
-	go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.27.0 // indirect
-	go.opentelemetry.io/otel/metric v1.33.0 // indirect
-	go.opentelemetry.io/otel/sdk v1.33.0 // indirect
-	go.opentelemetry.io/otel/sdk/metric v1.32.0 // indirect
-	go.opentelemetry.io/otel/trace v1.33.0 // indirect
-	go.opentelemetry.io/proto/otlp v1.4.0 // indirect
+	go.opentelemetry.io/contrib/bridges/prometheus v0.57.0 // indirect
+	go.opentelemetry.io/contrib/exporters/autoexport v0.57.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.60.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/net/http/httptrace/otelhttptrace v0.60.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.60.0 // indirect
+	go.opentelemetry.io/otel v1.36.0 // indirect
+	go.opentelemetry.io/otel/exporters/otlp/otlplog/otlploggrpc v0.8.0 // indirect
+	go.opentelemetry.io/otel/exporters/otlp/otlplog/otlploghttp v0.8.0 // indirect
+	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.35.0 // indirect
+	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp v1.35.0 // indirect
+	go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.35.0 // indirect
+	go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.35.0 // indirect
+	go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp v1.35.0 // indirect
+	go.opentelemetry.io/otel/exporters/prometheus v0.54.0 // indirect
+	go.opentelemetry.io/otel/exporters/stdout/stdoutlog v0.8.0 // indirect
+	go.opentelemetry.io/otel/exporters/stdout/stdoutmetric v1.32.0 // indirect
+	go.opentelemetry.io/otel/exporters/stdout/stdouttrace v1.32.0 // indirect
+	go.opentelemetry.io/otel/log v0.8.0 // indirect
+	go.opentelemetry.io/otel/metric v1.36.0 // indirect
+	go.opentelemetry.io/otel/sdk v1.36.0 // indirect
+	go.opentelemetry.io/otel/sdk/log v0.8.0 // indirect
+	go.opentelemetry.io/otel/sdk/metric v1.36.0 // indirect
+	go.opentelemetry.io/otel/trace v1.36.0 // indirect
+	go.opentelemetry.io/proto/otlp v1.5.0 // indirect
 	go.step.sm/cli-utils v0.9.0 // indirect
 	go.step.sm/crypto v0.45.0 // indirect
 	go.step.sm/linkedca v0.20.1 // indirect
 	go.uber.org/automaxprocs v1.5.3 // indirect
-	go.uber.org/mock v0.5.0 // indirect
+	go.uber.org/mock v0.6.0 // indirect
 	go.uber.org/multierr v1.11.0 // indirect
 	go.uber.org/zap/exp v0.2.0 // indirect
+	go.yaml.in/yaml/v3 v3.0.4 // indirect
 	golang.org/x/crypto/x509roots/fallback v0.0.0-20240507223354-67b13616a595 // indirect
-	golang.org/x/exp v0.0.0-20241215155358-4a5509556b9e // indirect
-	golang.org/x/mod v0.23.0 // indirect
-	golang.org/x/term v0.29.0 // indirect
-	golang.org/x/text v0.22.0 // indirect
-	golang.org/x/time v0.8.0 // indirect
-	golang.org/x/tools v0.30.0 // indirect
+	golang.org/x/exp v0.0.0-20250408133849-7e4ce0ab07d0 // indirect
+	golang.org/x/mod v0.27.0 // indirect
+	golang.org/x/oauth2 v0.30.0 // indirect
+	golang.org/x/text v0.28.0 // indirect
+	golang.org/x/time v0.11.0 // indirect
+	golang.org/x/tools v0.36.0 // indirect
 	golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2 // indirect
-	google.golang.org/genproto v0.0.0-20240401170217-c3f982113cda // indirect
-	google.golang.org/genproto/googleapis/api v0.0.0-20241209162323-e6fa225c2576 // indirect
+	google.golang.org/genproto/googleapis/api v0.0.0-20250528174236-200df99c418a // indirect
+	gopkg.in/evanphx/json-patch.v4 v4.12.0 // indirect
+	gopkg.in/inf.v0 v0.9.1 // indirect
+	gopkg.in/ini.v1 v1.67.0 // indirect
+	gopkg.in/yaml.v2 v2.4.0 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
 	gvisor.dev/gvisor v0.0.0-20230927004350-cbd86285d259 // indirect
 	howett.net/plist v1.0.0 // indirect
+	k8s.io/api v0.32.3 // indirect
+	k8s.io/apimachinery v0.32.3 // indirect
+	k8s.io/client-go v0.32.3 // indirect
+	k8s.io/klog/v2 v2.130.1 // indirect
+	k8s.io/kube-openapi v0.0.0-20241105132330-32ad38e42d3f // indirect
+	k8s.io/utils v0.0.0-20241104100929-3ea5e8cea738 // indirect
 	lukechampine.com/blake3 v1.3.0 // indirect
 	modernc.org/libc v1.61.13 // indirect
 	modernc.org/mathutil v1.7.1 // indirect
 	modernc.org/memory v1.8.2 // indirect
+	sigs.k8s.io/json v0.0.0-20241010143419-9aa6b5e7a4b3 // indirect
+	sigs.k8s.io/structured-merge-diff/v4 v4.4.2 // indirect
+	sigs.k8s.io/yaml v1.4.0 // indirect
+	tags.cncf.io/container-device-interface v1.0.1 // indirect
 )

internal/cli/build.go

@@ -2,175 +2,198 @@ package cli
 
 import (
 	"context"
-	"encoding/base64"
-	"encoding/json"
+	"errors"
 	"fmt"
-	"os"
+	"strings"
 
+	"github.com/charmbracelet/lipgloss"
 	composetypes "github.com/compose-spec/compose-go/v2/types"
-	"github.com/distribution/reference"
-	"github.com/docker/cli/cli/config"
-	"github.com/docker/docker/api/types"
-	"github.com/docker/docker/api/types/image"
-	dockerclient "github.com/docker/docker/client"
-	"github.com/docker/docker/pkg/archive"
-	"github.com/docker/docker/pkg/jsonmessage"
-	"github.com/docker/docker/registry"
-	"github.com/moby/term"
+	mapset "github.com/deckarep/golang-set/v2"
+	"github.com/docker/cli/cli/command"
+	"github.com/docker/cli/cli/flags"
+	composeapi "github.com/docker/compose/v2/pkg/api"
+	composev2 "github.com/docker/compose/v2/pkg/compose"
+	"github.com/docker/compose/v2/pkg/progress"
+	"github.com/psviderski/uncloud/pkg/client"
+	"github.com/psviderski/uncloud/pkg/client/compose"
 )
 
-type BuildOptions struct {
-	Files    []string
-	Profiles []string
-	Services []string
-	Push     bool
+// BuildServicesOptions contains options for building services in a Compose project.
+type BuildServicesOptions struct {
+	// BuildArgs sets build-time variables for services. Used in Dockerfiles that declare variables with ARG.
+	BuildArgs []string
+	// Check the build configuration for services without building them.
+	Check bool
+	// Deps enables to also build services declared as dependencies of the selected Services.
+	Deps bool
+	// NoCache disables the use of cache when building images.
 	NoCache bool
+	// Pull attempts to pull newer versions of the base images before building.
+	Pull bool
+	// Services specifies which services to build. If empty, all services with a build config are built.
+	Services []string
+
+	// Push targets are mutually exclusive.
+	// PushCluster uploads the built images to cluster machines after building.
+	PushCluster bool
+	// PushRegistry uploads the built images to external registries after building.
+	PushRegistry bool
+
+	// Cluster-specific options (only used if PushCluster is true).
+	// Machines is a list of machine names or IDs to push the image to. If empty, images are pushed to all machines.
+	Machines []string
 }
 
-// GetServicesThatNeedBuild returns a map of services that require building
-func GetServicesThatNeedBuild(project *composetypes.Project) map[string]composetypes.ServiceConfig {
-	servicesToBuild := make(map[string]composetypes.ServiceConfig, len(project.Services))
-	for serviceName, service := range project.Services {
-		if service.Build == nil {
+// BuildServices builds images for services in the Compose project.
+func (cli *CLI) BuildServices(ctx context.Context, project *composetypes.Project, opts BuildServicesOptions) error {
+	// Validate push targets.
+	if opts.PushCluster && opts.PushRegistry {
+		return fmt.Errorf("cannot specify both PushCluster and PushRegistry: choose one push target")
+	}
+
+	// Build service images using Compose implementation.
+	dockerCli, err := command.NewDockerCli()
+	if err != nil {
+		return fmt.Errorf("create docker client: %w", err)
+	}
+	// Initialise the Docker CLI with default options.
+	if err = dockerCli.Initialize(flags.NewClientOptions()); err != nil {
+		return fmt.Errorf("initialise docker client: %w", err)
+	}
+
+	composeService := composev2.NewComposeService(dockerCli)
+	buildOpts := composeapi.BuildOptions{
+		Args:     composetypes.NewMappingWithEquals(opts.BuildArgs),
+		Check:    opts.Check,
+		Deps:     opts.Deps,
+		NoCache:  opts.NoCache,
+		Pull:     opts.Pull,
+		Push:     opts.PushRegistry,
+		Services: opts.Services,
+	}
+
+	if err = composeService.Build(ctx, project, buildOpts); err != nil {
+		return err
+	}
+
+	if !opts.PushCluster {
+		return nil
+	}
+
+	// Push built service images to cluster machines.
+	builtServices, err := ServicesThatNeedBuild(project, opts.Services, opts.Deps)
+	if err != nil {
+		return fmt.Errorf("determine built services: %w", err)
+	}
+	if len(builtServices) == 0 {
+		// No services were built, nothing to push.
+		return nil
+	}
+
+	// Add a line break after the build output.
+	fmt.Fprintln(cli.ProgressOut())
+
+	clusterClient, err := cli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer clusterClient.Close()
+
+	// Push one service image at a time.
+	var errs []error
+	for _, s := range builtServices {
+		if s.Image == "" {
+			// Skip services without an image name (shouldn't happen for services with build config).
 			continue
 		}
-		servicesToBuild[serviceName] = service
-	}
-	return servicesToBuild
-}
 
-// BuildServices builds the services defined in the provided map.
-func BuildServices(ctx context.Context, servicesToBuild map[string]composetypes.ServiceConfig, opts BuildOptions) error {
-	fmt.Println("Building services...")
+		// Push to the specified machines falling back to service x-machines.
+		// If none specified, push to *all* cluster machines.
+		var pushOpts client.PushImageOptions
 
-	// Init docker client (can be local or remote, depending on DOCKER_HOST environment variable)
-	dockerCli, err := dockerclient.NewClientWithOpts(dockerclient.FromEnv, dockerclient.WithAPIVersionNegotiation())
-	if err != nil {
-		return err
-	}
-	defer dockerCli.Close()
-
-	serviceImages := make(map[string]string, len(servicesToBuild))
-
-	// Build the services using the local docker client and compose libraries
-	for _, service := range servicesToBuild {
-		fmt.Printf("Building service: %s\n", service.Name)
-		imageName, err := buildSingleService(ctx, dockerCli, service, opts)
-		if err != nil {
-			return fmt.Errorf("build service %s: %w", service.Name, err)
-		}
-		serviceImages[service.Name] = imageName
-	}
-	fmt.Printf("Service images are built.\n")
-
-	if opts.Push {
-		err = pushServiceImages(ctx, dockerCli, serviceImages)
+		if len(opts.Machines) > 0 {
+			pushOpts.Machines = opts.Machines
+		} else if machines, ok := s.Extensions[compose.MachinesExtensionKey].(compose.MachinesSource); ok {
+			pushOpts.Machines = machines
 		}
 
-	return err
-}
-
-// buildSingleService builds a single service using the Docker client and Compose libraries.
-func buildSingleService(ctx context.Context, dockerCli *dockerclient.Client, service composetypes.ServiceConfig, opts BuildOptions) (string, error) {
-	if service.Build == nil {
-		return "", fmt.Errorf("service %s has no build configuration", service.Name)
-	}
-	if service.Image == "" {
-		return "", fmt.Errorf("service %s has no image specified; building services without image is not supported yet", service.Name)
+		if len(pushOpts.Machines) == 0 {
+			pushOpts.AllMachines = true
 		}
 
-	buildContextPath := service.Build.Context
-	imageName := service.Image
-
-	// Create a tar archive of the build context
-	buildContext, err := archive.TarWithOptions(buildContextPath, &archive.TarOptions{})
-	if err != nil {
-		return "", fmt.Errorf("failed to create build context for service %s: %w", service.Name, err)
-	}
-
-	buildOptions := types.ImageBuildOptions{
-		// TODO: Support Dockerfiles outside the build context
-		// See https://github.com/docker/compose/blob/cf89fd1aa1328d5af77658ccc5a1e1b29981ae80/pkg/compose/build_classic.go#L92
-		Dockerfile: service.Build.Dockerfile,
-		Tags:       []string{imageName},
-		Remove:     true, // Remove intermediate containers
-		NoCache:    opts.NoCache,
-	}
-
-	buildResponse, err := dockerCli.ImageBuild(ctx, buildContext, buildOptions)
-	if err != nil {
-		return "", fmt.Errorf("failed to build image for service %s: %w", service.Name, err)
-	}
-	defer buildResponse.Body.Close()
-
-	// Display the build response
-	fd, isTerminal := term.GetFdInfo(os.Stdout)
-	if err := jsonmessage.DisplayJSONMessagesStream(buildResponse.Body, os.Stdout, fd, isTerminal, nil); err != nil {
-		return "", fmt.Errorf("failed to display build response for service %s: %w", service.Name, err)
-	}
-
-	return imageName, nil
-}
-
-// pushSingleServiceImage pushes a single service image.
-func pushSingleServiceImage(ctx context.Context, dockerCli *dockerclient.Client, serviceName string, imageName string) error {
-	ref, err := reference.ParseNormalizedNamed(imageName)
-	if err != nil {
-		return err
-	}
-
-	repoInfo, err := registry.ParseRepositoryInfo(ref)
-	if err != nil {
-		return err
-	}
-
-	registryKey := repoInfo.Index.Name
-	if repoInfo.Index.Official {
-		registryKey = registry.IndexServer
-	}
-
-	// Load the Docker config file with auth details, if available
-	configFile := config.LoadDefaultConfigFile(os.Stderr)
-
-	authConfig, err := configFile.GetAuthConfig(registryKey)
-	if err != nil {
-		return err
-	}
-
-	authJSON, err := json.Marshal(authConfig)
-	if err != nil {
-		return fmt.Errorf("failed to marshal auth config for registry %s: %w", registryKey, err)
-	}
-
-	authStr := base64.URLEncoding.EncodeToString(authJSON)
-
-	pushOptions := image.PushOptions{
-		RegistryAuth: authStr,
-	}
-	pushResponse, err := dockerCli.ImagePush(ctx, imageName, pushOptions)
-	if err != nil {
-		return fmt.Errorf("failed to push image %s: %w", imageName, err)
-	}
-	defer pushResponse.Close()
-
-	fmt.Printf("Pushing image %s for service %s...\n", imageName, serviceName)
-
-	fd, isTerminal := term.GetFdInfo(os.Stdout)
-	if err := jsonmessage.DisplayJSONMessagesStream(pushResponse, os.Stdout, fd, isTerminal, nil); err != nil {
-		return fmt.Errorf("failed to display push response for image %s: %w", imageName, err)
-	}
-
-	fmt.Printf("Image %s pushed successfully.\n", imageName)
-	return nil
-}
-
-// pushServiceImages pushes all built service images to the registry.
-func pushServiceImages(ctx context.Context, dockerCli *dockerclient.Client, serviceImages map[string]string) error {
-	fmt.Printf("Pushing images...\n")
-	for serviceName, imageName := range serviceImages {
-		if err := pushSingleServiceImage(ctx, dockerCli, serviceName, imageName); err != nil {
-			return fmt.Errorf("push image for service %s: %w", serviceName, err)
-		}
+		boldStyle := lipgloss.NewStyle().Bold(true)
+		err = progress.RunWithTitle(ctx, func(ctx context.Context) error {
+			if err = clusterClient.PushImage(ctx, s.Image, pushOpts); err != nil {
+				return fmt.Errorf("push image '%s' for service '%s': %w", s.Image, s.Name, err)
 			}
 			return nil
+		}, cli.ProgressOut(), fmt.Sprintf("Pushing image %s to cluster", boldStyle.Render(s.Image)))
+		// Collect errors to try pushing all images.
+		if err != nil {
+			errs = append(errs, err)
+		}
+	}
+
+	return errors.Join(errs...)
+}
+
+// ServicesThatNeedBuild returns a list of services that require building.
+// deps indicates whether to include services that are dependencies of the selected services.
+// Implementation is based on the logic from docker/compose/v2/pkg/compose/build.go.
+func ServicesThatNeedBuild(
+	project *composetypes.Project, selectedServices []string, deps bool,
+) ([]composetypes.ServiceConfig, error) {
+	servicesToBuild := make([]composetypes.ServiceConfig, 0, len(project.Services))
+
+	var policy composetypes.DependencyOption = composetypes.IgnoreDependencies
+	if deps {
+		policy = composetypes.IncludeDependencies
+	}
+
+	// Also include services used as build.additional_contexts with service: prefix.
+	selectedServices = includeAdditionalContextsServices(project, selectedServices)
+	// Some build dependencies we just introduced may not be enabled, enable them.
+	var err error
+	if project, err = project.WithServicesEnabled(selectedServices...); err != nil {
+		return nil, err
+	}
+	if project, err = project.WithSelectedServices(selectedServices); err != nil {
+		return nil, err
+	}
+
+	err = project.ForEachService(selectedServices, func(serviceName string, service *composetypes.ServiceConfig) error {
+		if service.Build != nil {
+			servicesToBuild = append(servicesToBuild, *service)
+		}
+		return nil
+	}, policy)
+	if err != nil {
+		return nil, err
+	}
+
+	return servicesToBuild, nil
+}
+
+// includeAdditionalContextsServices adds services referenced in build.additional_contexts to the list
+// of selected services.
+func includeAdditionalContextsServices(project *composetypes.Project, selectedServices []string) []string {
+	servicesWithDependencies := mapset.NewSet(selectedServices...)
+	for _, service := range selectedServices {
+		s, ok := project.Services[service]
+		if !ok {
+			s = project.DisabledServices[service]
+		}
+		if s.Build != nil {
+			for _, target := range s.Build.AdditionalContexts {
+				if name, found := strings.CutPrefix(target, composetypes.ServicePrefix); found {
+					servicesWithDependencies.Add(name)
+				}
+			}
+		}
+	}
+	if servicesWithDependencies.Cardinality() > len(selectedServices) {
+		return includeAdditionalContextsServices(project, servicesWithDependencies.ToSlice())
+	}
+
+	return servicesWithDependencies.ToSlice()
 }

internal/cli/cli.go

@@ -2,7 +2,6 @@ package cli
 
 import (
 	"context"
-	"errors"
 	"fmt"
 	"net/netip"
 	"os"
@@ -10,7 +9,6 @@ import (
 
 	"github.com/docker/cli/cli/streams"
 	"github.com/psviderski/uncloud/internal/cli/config"
-	"github.com/psviderski/uncloud/internal/fs"
 	"github.com/psviderski/uncloud/internal/machine"
 	"github.com/psviderski/uncloud/internal/machine/api/pb"
 	"github.com/psviderski/uncloud/internal/sshexec"
@@ -32,12 +30,14 @@ const (
 type CLI struct {
 	Config          *config.Config
 	conn            *config.MachineConnection
+	contextOverride string
 }
 
 // New creates a new CLI instance with the given config path or remote machine connection.
 // If the connection is provided, the config is ignored for all operations which is useful for interacting with
 // a cluster without creating a config.
-func New(configPath string, conn *config.MachineConnection) (*CLI, error) {
+// If a non-empty context name is given, it will override the current default.
+func New(configPath string, conn *config.MachineConnection, contextName string) (*CLI, error) {
 	if conn != nil {
 		return &CLI{conn: conn}, nil
 	}
@@ -49,6 +49,7 @@ func New(configPath string, conn *config.MachineConnection) (*CLI, error) {
 
 	return &CLI{
 		Config:          cfg,
+		contextOverride: contextName,
 	}, nil
 }
 
@@ -70,11 +71,30 @@ func (cli *CLI) SetCurrentContext(name string) error {
 	return cli.Config.Save()
 }
 
-// ConnectCluster connects to a cluster using the given context name or the current context if not specified.
+func (cli *CLI) GetContextOverrideOrCurrent() string {
+	contextName := cli.contextOverride
+	if contextName == "" {
+		contextName = cli.Config.CurrentContext
+	}
+	return contextName
+}
+
+// ConnectCluster connects to a cluster using the context override or the current context if not specified.
 // If the CLI was initialised with a machine connection, the config is ignored and the connection is used instead.
-func (cli *CLI) ConnectCluster(ctx context.Context, contextName string) (*client.Client, error) {
+func (cli *CLI) ConnectCluster(ctx context.Context) (*client.Client, error) {
+	return cli.ConnectClusterWithOptions(ctx, ConnectOptions{
+		// Default to showing progress for CLI usage.
+		ShowProgress: true,
+	})
+}
+
+// ConnectClusterWithOptions connects to a cluster with the given options.
+// If the CLI was initialised with a machine connection, the config is ignored and the connection is used instead.
+// If the CLI has an override context, it is used instead of the current default.
+// Options are useful when using the CLI as a library where you may want to disable visual feedback.
+func (cli *CLI) ConnectClusterWithOptions(ctx context.Context, opts ConnectOptions) (*client.Client, error) {
 	if cli.conn != nil {
-		return connectCluster(ctx, *cli.conn)
+		return ConnectCluster(ctx, *cli.conn, opts)
 	}
 
 	if len(cli.Config.Contexts) == 0 {
@@ -84,6 +104,8 @@ func (cli *CLI) ConnectCluster(ctx context.Context, contextName string) (*client
 			cli.Config.Path(),
 		)
 	}
+
+	contextName := cli.contextOverride
 	if contextName == "" {
 		// If the cluster is not specified, use the current cluster if set.
 		if cli.Config.CurrentContext == "" {
@@ -116,38 +138,20 @@ func (cli *CLI) ConnectCluster(ctx context.Context, contextName string) (*client
 		)
 	}
 
-	// TODO: iterate over all connections and try to connect to the cluster using the first successful connection.
-	conn := cfg.Connections[0]
-
-	c, err := connectCluster(ctx, conn)
-	if err != nil {
-		return nil, fmt.Errorf("connect to cluster (context '%s'): %w", contextName, err)
-	}
-
+	// Try each connection in order until one succeeds.
+	var lastErr error
+	for _, conn := range cfg.Connections {
+		c, err := ConnectCluster(ctx, conn, opts)
+		if err == nil {
 			return c, nil
-}
-
-func connectCluster(ctx context.Context, conn config.MachineConnection) (*client.Client, error) {
-	if conn.SSH != "" {
-		user, host, port, err := conn.SSH.Parse()
-		if err != nil {
-			return nil, fmt.Errorf("parse SSH connection %q: %w", conn.SSH, err)
 		}
 
-		keyPath := fs.ExpandHomeDir(conn.SSHKeyFile)
-
-		sshConfig := &connector.SSHConnectorConfig{
-			User:    user,
-			Host:    host,
-			Port:    port,
-			KeyPath: keyPath,
-		}
-		return client.New(ctx, connector.NewSSHConnector(sshConfig))
-	} else if conn.TCP != nil && conn.TCP.IsValid() {
-		return client.New(ctx, connector.NewTCPConnector(*conn.TCP))
+		lastErr = err
 	}
 
-	return nil, errors.New("connection configuration is invalid")
+	return nil, fmt.Errorf("failed to connect to cluster context '%s': "+
+		"all connections (%d) in the Uncloud config (%s) failed; last error: %w",
+		contextName, len(cfg.Connections), cli.Config.Path(), lastErr)
 }
 
 type InitClusterOptions struct {
@@ -156,6 +160,7 @@ type InitClusterOptions struct {
 	Network       netip.Prefix
 	PublicIP      *netip.Addr
 	RemoteMachine *RemoteMachine
+	SkipInstall   bool
 	Version       string
 }
 
@@ -166,7 +171,7 @@ func (cli *CLI) InitCluster(ctx context.Context, opts InitClusterOptions) (*clie
 		return cli.initRemoteMachine(ctx, opts)
 	}
 	// TODO: implement local machine initialisation
-	return nil, fmt.Errorf("local machine initialisation is not implemented yet")
+	return nil, fmt.Errorf("local machine initialisation is not implemented yet. Please specify a remote machine")
 }
 
 func (cli *CLI) initRemoteMachine(ctx context.Context, opts InitClusterOptions) (*client.Client, error) {
@@ -175,7 +180,7 @@ func (cli *CLI) initRemoteMachine(ctx context.Context, opts InitClusterOptions)
 		return nil, err
 	}
 
-	machineClient, err := provisionRemoteMachine(ctx, opts.RemoteMachine, opts.Version)
+	machineClient, err := provisionOrConnectRemoteMachine(ctx, opts.RemoteMachine, opts.SkipInstall, opts.Version)
 	if err != nil {
 		return nil, err
 	}
@@ -237,8 +242,21 @@ func (cli *CLI) initRemoteMachine(ctx context.Context, opts InitClusterOptions)
 
 	// Save the machine's SSH connection details in the context config.
 	connCfg := config.MachineConnection{
-		SSH:        config.NewSSHDestination(opts.RemoteMachine.User, opts.RemoteMachine.Host, opts.RemoteMachine.Port),
 		SSHKeyFile: opts.RemoteMachine.KeyPath,
+		MachineID:  resp.Machine.Id,
+	}
+	if opts.RemoteMachine.UseSSHCLI {
+		connCfg.SSHCLI = config.NewSSHDestination(
+			opts.RemoteMachine.User,
+			opts.RemoteMachine.Host,
+			opts.RemoteMachine.Port,
+		)
+	} else {
+		connCfg.SSH = config.NewSSHDestination(
+			opts.RemoteMachine.User,
+			opts.RemoteMachine.Host,
+			opts.RemoteMachine.Port,
+		)
 	}
 	cli.Config.Contexts[contextName].Connections = append(cli.Config.Contexts[contextName].Connections, connCfg)
 	if err = cli.Config.Save(); err != nil {
@@ -274,10 +292,10 @@ func (cli *CLI) newContextName(name string) (string, error) {
 }
 
 type AddMachineOptions struct {
-	Context       string
 	MachineName   string
 	PublicIP      *netip.Addr
 	RemoteMachine *RemoteMachine
+	SkipInstall   bool
 	Version       string
 }
 
@@ -286,11 +304,8 @@ type AddMachineOptions struct {
 // cluster. The machine client is connected to the new machine and can be used to interact with it.
 // Both client should be closed after use by the caller.
 func (cli *CLI) AddMachine(ctx context.Context, opts AddMachineOptions) (*client.Client, *client.Client, error) {
-	contextName := opts.Context
-	if contextName == "" {
-		contextName = cli.Config.CurrentContext
-	}
-	c, err := cli.ConnectCluster(ctx, contextName)
+	contextName := cli.GetContextOverrideOrCurrent()
+	c, err := cli.ConnectCluster(ctx)
 	if err != nil {
 		return nil, nil, fmt.Errorf("connect to cluster (context '%s'): %w", contextName, err)
 	}
@@ -300,7 +315,7 @@ func (cli *CLI) AddMachine(ctx context.Context, opts AddMachineOptions) (*client
 		}
 	}()
 
-	machineClient, err := provisionRemoteMachine(ctx, opts.RemoteMachine, opts.Version)
+	machineClient, err := provisionOrConnectRemoteMachine(ctx, opts.RemoteMachine, opts.SkipInstall, opts.Version)
 	if err != nil {
 		return nil, nil, err
 	}
@@ -404,11 +419,21 @@ func (cli *CLI) AddMachine(ctx context.Context, opts AddMachineOptions) (*client
 
 	// Save the machine's SSH connection details in the context config.
 	connCfg := config.MachineConnection{
-		SSH:        config.NewSSHDestination(opts.RemoteMachine.User, opts.RemoteMachine.Host, opts.RemoteMachine.Port),
 		SSHKeyFile: opts.RemoteMachine.KeyPath,
+		MachineID:  addResp.Machine.Id,
 	}
-	if contextName == "" {
-		contextName = cli.Config.CurrentContext
+	if opts.RemoteMachine.UseSSHCLI {
+		connCfg.SSHCLI = config.NewSSHDestination(
+			opts.RemoteMachine.User,
+			opts.RemoteMachine.Host,
+			opts.RemoteMachine.Port,
+		)
+	} else {
+		connCfg.SSH = config.NewSSHDestination(
+			opts.RemoteMachine.User,
+			opts.RemoteMachine.Host,
+			opts.RemoteMachine.Port,
+		)
 	}
 	cli.Config.Contexts[contextName].Connections = append(cli.Config.Contexts[contextName].Connections, connCfg)
 	if err = cli.Config.Save(); err != nil {
@@ -418,15 +443,45 @@ func (cli *CLI) AddMachine(ctx context.Context, opts AddMachineOptions) (*client
 	return c, machineClient, nil
 }
 
-// provisionRemoteMachine installs the Uncloud daemon and dependencies on the remote machine over SSH and returns
-// a machine API client to interact with the machine. The client should be closed after use by the caller.
+// provisionOrConnectRemoteMachine installs the Uncloud daemon and dependencies on the remote machine over SSH and
+// returns a machine API client to interact with the machine. The client should be closed after use by the caller.
 // The version parameter specifies the version of the Uncloud daemon to install. If empty, the latest version is used.
+// If skipInstall is true, the installation step is skipped, and it is assumed that the Uncloud daemon and dependencies
+// are already installed and running.
 // The remoteMachine.SSHKeyPath could be updated to the default SSH key path if it is not set and the SSH agent
 // authentication fails.
-func provisionRemoteMachine(
-	ctx context.Context, remoteMachine *RemoteMachine, version string,
+func provisionOrConnectRemoteMachine(
+	ctx context.Context, remoteMachine *RemoteMachine, skipInstall bool, version string,
 ) (*client.Client, error) {
-	// Provision the remote machine by installing the Uncloud daemon and dependencies over SSH.
+	// Use SSH CLI
+	if remoteMachine.UseSSHCLI {
+		exec := sshexec.NewSSHCLIRemote(
+			remoteMachine.User,
+			remoteMachine.Host,
+			remoteMachine.Port,
+			remoteMachine.KeyPath,
+		)
+
+		if !skipInstall {
+			if err := provisionMachine(ctx, exec, version); err != nil {
+				return nil, fmt.Errorf("provision machine: %w", err)
+			}
+		}
+
+		sshConfig := &connector.SSHConnectorConfig{
+			User:    remoteMachine.User,
+			Host:    remoteMachine.Host,
+			Port:    remoteMachine.Port,
+			KeyPath: remoteMachine.KeyPath,
+		}
+		machineClient, err := client.New(ctx, connector.NewSSHCLIConnector(sshConfig))
+		if err != nil {
+			return nil, fmt.Errorf("connect to remote machine: %w", err)
+		}
+		return machineClient, nil
+	}
+
+	// Use Go SSH
 	sshClient, err := sshexec.Connect(remoteMachine.User, remoteMachine.Host, remoteMachine.Port, remoteMachine.KeyPath)
 	// If the SSH connection using SSH agent fails and no key path is provided, try to use the default SSH key.
 	if err != nil && remoteMachine.KeyPath == "" {
@@ -441,14 +496,17 @@ func provisionRemoteMachine(
 			config.NewSSHDestination(remoteMachine.User, remoteMachine.Host, remoteMachine.Port), err,
 		)
 	}
+
+	if !skipInstall {
+		// Provision the remote machine by installing the Uncloud daemon and dependencies over SSH.
 		exec := sshexec.NewRemote(sshClient)
-	// Install and run the Uncloud daemon and dependencies on the remote machine.
 		if err = provisionMachine(ctx, exec, version); err != nil {
 			return nil, fmt.Errorf("provision machine: %w", err)
 		}
+	}
 
 	var machineClient *client.Client
-	if remoteMachine.User == "root" {
+	if remoteMachine.User == "root" || skipInstall {
 		// Create a machine API client over the established SSH connection to the remote machine.
 		machineClient, err = client.New(ctx, connector.NewSSHConnectorFromClient(sshClient))
 	} else {

internal/cli/config/connection.go

@@ -1,6 +1,8 @@
 package config
 
 import (
+	"errors"
+	"fmt"
 	"net"
 	"net/netip"
 	"strconv"
@@ -16,12 +18,54 @@ const (
 
 type MachineConnection struct {
 	SSH        SSHDestination `yaml:"ssh,omitempty"`
+	SSHCLI     SSHDestination `yaml:"ssh_cli,omitempty"`
 	SSHKeyFile string         `yaml:"ssh_key_file,omitempty"`
 	// TCP is the address and port of the machine's API server.
 	// The pointer is used to omit the field when not set. Otherwise, yaml marshalling includes an empty object.
 	TCP *netip.AddrPort `yaml:"tcp,omitempty"`
+	// Unix is the path to the machine's API unix socket.
+	Unix      string        `yaml:"unix,omitempty"`
 	Host      string        `yaml:"host,omitempty"`
 	PublicKey secret.Secret `yaml:"public_key,omitempty"`
+	MachineID string        `yaml:"machine_id,omitempty"`
+}
+
+func (c MachineConnection) String() string {
+	if c.SSH != "" {
+		return "ssh://" + string(c.SSH)
+	} else if c.SSHCLI != "" {
+		return "ssh+cli://" + string(c.SSHCLI)
+	} else if c.TCP != nil && c.TCP.IsValid() {
+		return fmt.Sprintf("tcp://%s", c.TCP)
+	} else if c.Unix != "" {
+		return fmt.Sprintf("unix://%s", c.Unix)
+	}
+	return "unknown connection"
+}
+
+func (c *MachineConnection) Validate() error {
+	setCount := 0
+	if c.SSH != "" {
+		setCount++
+	}
+	if c.SSHCLI != "" {
+		setCount++
+	}
+	if c.TCP != nil && c.TCP.IsValid() {
+		setCount++
+	}
+	if c.Unix != "" {
+		setCount++
+	}
+
+	if setCount == 0 {
+		return errors.New("no connection method specified (ssh, ssh_cli, tcp, or unix required)")
+	}
+	if setCount > 1 {
+		return errors.New("only one connection method allowed per connection (ssh, ssh_cli, tcp, or unix)")
+	}
+
+	return nil
 }
 
 // SSHDestination represents an SSH destination string in the canonical form of "user@host:port".

internal/cli/config/connection_test.go

@@ -0,0 +1,205 @@
+package config
+
+import (
+	"net/netip"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestMachineConnection_String(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name string
+		conn MachineConnection
+		want string
+	}{
+		{
+			name: "ssh connection",
+			conn: MachineConnection{
+				SSH: "user@host.com",
+			},
+			want: "ssh://user@host.com",
+		},
+		{
+			name: "ssh connection with port",
+			conn: MachineConnection{
+				SSH: "user@host.com:2222",
+			},
+			want: "ssh://user@host.com:2222",
+		},
+		{
+			name: "ssh_cli connection",
+			conn: MachineConnection{
+				SSHCLI: "user@host.com",
+			},
+			want: "ssh+cli://user@host.com",
+		},
+		{
+			name: "ssh_cli connection with port",
+			conn: MachineConnection{
+				SSHCLI: "user@host.com:2222",
+			},
+			want: "ssh+cli://user@host.com:2222",
+		},
+		{
+			name: "tcp connection",
+			conn: MachineConnection{
+				TCP: func() *netip.AddrPort {
+					addr := netip.MustParseAddrPort("10.0.0.1:8080")
+					return &addr
+				}(),
+			},
+			want: "tcp://10.0.0.1:8080",
+		},
+		{
+			name: "unix connection",
+			conn: MachineConnection{
+				Unix: "/run/uncloud/uncloud.sock",
+			},
+			want: "unix:///run/uncloud/uncloud.sock",
+		},
+		{
+			name: "no connection",
+			conn: MachineConnection{},
+			want: "unknown connection",
+		},
+		{
+			name: "ssh connection with machine id",
+			conn: MachineConnection{
+				SSH:       "user@host.com",
+				MachineID: "ed98b9f7575308c340263cd279e3b568",
+			},
+			// String() does not use MachineID; just verifying entry with it is valid
+			want: "ssh://user@host.com",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			got := tt.conn.String()
+			assert.Equal(t, tt.want, got)
+		})
+	}
+}
+
+func TestMachineConnection_Validate(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name    string
+		conn    MachineConnection
+		wantErr bool
+		errMsg  string
+	}{
+		{
+			name: "ssh only - valid",
+			conn: MachineConnection{
+				SSH: "user@host",
+			},
+			wantErr: false,
+		},
+		{
+			name: "ssh_cli only - valid",
+			conn: MachineConnection{
+				SSHCLI: "user@host",
+			},
+			wantErr: false,
+		},
+		{
+			name: "tcp only - valid",
+			conn: MachineConnection{
+				TCP: func() *netip.AddrPort {
+					addr := netip.MustParseAddrPort("10.0.0.1:8080")
+					return &addr
+				}(),
+			},
+			wantErr: false,
+		},
+		{
+			name: "unix only - valid",
+			conn: MachineConnection{
+				Unix: "/path/to/socket",
+			},
+			wantErr: false,
+		},
+		{
+			name:    "no connection method - error",
+			conn:    MachineConnection{},
+			wantErr: true,
+			errMsg:  "no connection method specified",
+		},
+		{
+			name: "ssh and ssh_cli - error",
+			conn: MachineConnection{
+				SSH:    "user@host",
+				SSHCLI: "user@host",
+			},
+			wantErr: true,
+			errMsg:  "only one connection method allowed",
+		},
+		{
+			name: "ssh and unix - error",
+			conn: MachineConnection{
+				SSH:  "user@host",
+				Unix: "/path/to/socket",
+			},
+			wantErr: true,
+			errMsg:  "only one connection method allowed",
+		},
+		{
+			name: "ssh and tcp - error",
+			conn: MachineConnection{
+				SSH: "user@host",
+				TCP: func() *netip.AddrPort {
+					addr := netip.MustParseAddrPort("10.0.0.1:8080")
+					return &addr
+				}(),
+			},
+			wantErr: true,
+			errMsg:  "only one connection method allowed",
+		},
+		{
+			name: "ssh_cli and tcp - error",
+			conn: MachineConnection{
+				SSHCLI: "user@host",
+				TCP: func() *netip.AddrPort {
+					addr := netip.MustParseAddrPort("10.0.0.1:8080")
+					return &addr
+				}(),
+			},
+			wantErr: true,
+			errMsg:  "only one connection method allowed",
+		},
+		{
+			name: "all three - error",
+			conn: MachineConnection{
+				SSH:    "user@host",
+				SSHCLI: "user@host2",
+				TCP: func() *netip.AddrPort {
+					addr := netip.MustParseAddrPort("10.0.0.1:8080")
+					return &addr
+				}(),
+			},
+			wantErr: true,
+			errMsg:  "only one connection method allowed",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			err := tt.conn.Validate()
+			if tt.wantErr {
+				assert.Error(t, err)
+				assert.Contains(t, err.Error(), tt.errMsg)
+			} else {
+				assert.NoError(t, err)
+			}
+		})
+	}
+}

internal/cli/config/context.go

@@ -4,3 +4,21 @@ type Context struct {
 	Name        string              `yaml:"-"`
 	Connections []MachineConnection `yaml:"connections"`
 }
+
+func (c *Context) SetDefaultConnection(index int) {
+	// Do nothing if the index is out of range.
+	if index < 0 || index >= len(c.Connections) {
+		return
+	}
+
+	selected := c.Connections[index]
+	newConnections := make([]MachineConnection, 0, len(c.Connections))
+	newConnections = append(newConnections, selected)
+
+	for i, conn := range c.Connections {
+		if i != index {
+			newConnections = append(newConnections, conn)
+		}
+	}
+	c.Connections = newConnections
+}

internal/cli/connect.go

@@ -0,0 +1,212 @@
+package cli
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"os"
+	"time"
+
+	"github.com/charmbracelet/bubbles/spinner"
+	tea "github.com/charmbracelet/bubbletea"
+	"github.com/charmbracelet/lipgloss"
+	"github.com/psviderski/uncloud/internal/cli/config"
+	"github.com/psviderski/uncloud/internal/fs"
+	"github.com/psviderski/uncloud/pkg/client"
+	"github.com/psviderski/uncloud/pkg/client/connector"
+)
+
+// ConnectOptions configures the behavior of cluster connection attempts.
+type ConnectOptions struct {
+	// Whether to show connection progress spinner if stdout is a terminal or progress logs if not.
+	ShowProgress bool
+}
+
+func ConnectCluster(ctx context.Context, conn config.MachineConnection, opts ConnectOptions) (*client.Client, error) {
+	if opts.ShowProgress {
+		return connectClusterWithProgress(ctx, conn)
+	}
+	return connectCluster(ctx, conn)
+}
+
+// connectClusterWithProgress connects to the cluster while displaying a progress spinner.
+// If the stdout is not a terminal, it falls back to simple progress logs to stderr.
+func connectClusterWithProgress(ctx context.Context, conn config.MachineConnection) (*client.Client, error) {
+	// If stdout is not a terminal, fall back to simple progress logs.
+	if !IsStdoutTerminal() {
+		fmt.Fprintln(os.Stderr, "Connecting to", conn.String())
+		cli, err := connectCluster(ctx, conn)
+		if err != nil {
+			fmt.Fprintln(os.Stderr, "Connection failed:", err)
+		} else {
+			fmt.Fprintln(os.Stderr, "Connected to cluster.")
+		}
+		return cli, err
+	}
+
+	// Run the connection TUI model.
+	p := tea.NewProgram(newConnectModel(ctx, conn))
+	model, err := p.Run()
+	if err != nil {
+		return nil, fmt.Errorf("run connection TUI: %w", err)
+	}
+
+	m := model.(connectModel)
+	return m.result.client, m.result.err
+}
+
+func connectCluster(ctx context.Context, conn config.MachineConnection) (*client.Client, error) {
+	// Determine which SSH type is configured
+	var sshDest config.SSHDestination
+	var useSSHCLI bool
+
+	// Validate connection configuration early to provide clear error messages.
+	if err := conn.Validate(); err != nil {
+		return nil, fmt.Errorf("invalid connection configuration: %w", err)
+	}
+
+	if conn.SSH != "" {
+		sshDest = conn.SSH
+		useSSHCLI = false
+	} else if conn.SSHCLI != "" {
+		sshDest = conn.SSHCLI
+		useSSHCLI = true
+	} else if conn.TCP != nil && conn.TCP.IsValid() {
+		return client.New(ctx, connector.NewTCPConnector(*conn.TCP))
+	} else if conn.Unix != "" {
+		return client.New(ctx, connector.NewUnixConnector(conn.Unix))
+	} else {
+		return nil, errors.New("connection configuration is invalid")
+	}
+
+	// Parse SSH destination and create config (shared for both types)
+	user, host, port, err := sshDest.Parse()
+	if err != nil {
+		return nil, fmt.Errorf("parse SSH connection %q: %w", sshDest, err)
+	}
+
+	keyPath := fs.ExpandHomeDir(conn.SSHKeyFile)
+
+	sshConfig := &connector.SSHConnectorConfig{
+		User:    user,
+		Host:    host,
+		Port:    port,
+		KeyPath: keyPath,
+	}
+
+	// Create appropriate connector based on type
+	if useSSHCLI {
+		return client.New(ctx, connector.NewSSHCLIConnector(sshConfig))
+	}
+	return client.New(ctx, connector.NewSSHConnector(sshConfig))
+}
+
+// connectModel is a TUI model for connecting to a cluster with a progress spinner.
+type connectModel struct {
+	ctx     context.Context
+	conn    config.MachineConnection
+	spinner spinner.Model
+	// showSpinner controls whether the spinner is visible (delayed to avoid flashing).
+	showSpinner bool
+	// done indicates whether the connection attempt has completed (successfully or with error).
+	done bool
+	// result holds the result of the connection attempt.
+	result connectResultMsg
+}
+
+type connectResultMsg struct {
+	client *client.Client
+	err    error
+}
+
+// showSpinnerMsg is sent after a delay to show the spinner.
+type showSpinnerMsg struct{}
+
+func newConnectModel(ctx context.Context, conn config.MachineConnection) connectModel {
+	s := spinner.New()
+	s.Spinner = spinner.MiniDot
+	s.Style = lipgloss.NewStyle().Foreground(lipgloss.Color("3")) // the same yellow as in compose progress
+
+	return connectModel{
+		ctx:     ctx,
+		conn:    conn,
+		spinner: s,
+	}
+}
+
+func (m connectModel) Init() tea.Cmd {
+	return tea.Batch(
+		m.spinner.Tick,
+		m.connect(),
+		m.delayShowSpinner(),
+	)
+}
+
+func (m connectModel) connect() tea.Cmd {
+	return func() tea.Msg {
+		cli, err := connectCluster(m.ctx, m.conn)
+		return connectResultMsg{
+			client: cli,
+			err:    err,
+		}
+	}
+}
+
+// delayShowSpinner returns a command that sends a message to show the spinner after a delay.
+// This avoids flashing the spinner if the connection is fast.
+func (m connectModel) delayShowSpinner() tea.Cmd {
+	return tea.Tick(500*time.Millisecond, func(t time.Time) tea.Msg {
+		return showSpinnerMsg{}
+	})
+}
+
+func (m connectModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
+	select {
+	case <-m.ctx.Done():
+		m.result.err = m.ctx.Err()
+		m.done = true
+		return m, tea.Quit
+	default:
+	}
+
+	switch msg := msg.(type) {
+	case connectResultMsg:
+		m.result = msg
+		m.done = true
+		return m, tea.Quit
+
+	case showSpinnerMsg:
+		// Only show spinner if connection hasn't completed yet.
+		if !m.done {
+			m.showSpinner = true
+		}
+		return m, nil
+
+	case spinner.TickMsg:
+		var cmd tea.Cmd
+		m.spinner, cmd = m.spinner.Update(msg)
+		return m, cmd
+
+	case tea.KeyMsg:
+		if msg.Type == tea.KeyCtrlC {
+			m.result.err = fmt.Errorf("connection cancelled")
+			m.done = true
+			return m, tea.Quit
+		}
+	}
+
+	return m, nil
+}
+
+func (m connectModel) View() string {
+	// Don't show anything if done or spinner not yet visible.
+	if m.done || !m.showSpinner {
+		return ""
+	}
+
+	style := lipgloss.NewStyle().Foreground(lipgloss.Color("153"))
+	return fmt.Sprintf("%s %s\n",
+		m.spinner.View(),
+		fmt.Sprintf("Connecting to %s", style.Render(m.conn.String())),
+	)
+}

internal/cli/flags.go

@@ -1,7 +1,11 @@
 package cli
 
 import (
+	"log"
+	"os"
 	"strings"
+
+	"github.com/spf13/cobra"
 )
 
 // ExpandCommaSeparatedValues takes a slice of strings and expands any comma-separated values into individual elements.
@@ -21,3 +25,12 @@ func ExpandCommaSeparatedValues(values []string) []string {
 
 	return expanded
 }
+
+// BindEnvToFlag assigns the value of an environment variable to the given command flag if the flag has not been set.
+func BindEnvToFlag(cmd *cobra.Command, flagName, envVar string) {
+	if value := os.Getenv(envVar); value != "" && !cmd.Flags().Changed(flagName) {
+		if err := cmd.Flags().Set(flagName, value); err != nil {
+			log.Fatalf("Failed to bind environment variable '%s' to flag '%s': %v", envVar, flagName, err)
+		}
+	}
+}

internal/cli/machine.go

@@ -25,6 +25,7 @@ type RemoteMachine struct {
 	Host      string
 	Port      int
 	KeyPath   string
+	UseSSHCLI bool // indicates ssh+cli:// should be used
 }
 
 func installCmd(user string, version string) string {

internal/cli/prompt.go

@@ -1,6 +1,11 @@
 package cli
 
-import "github.com/charmbracelet/huh"
+import (
+	"os"
+
+	"github.com/charmbracelet/huh"
+	"golang.org/x/term"
+)
 
 func Confirm() (bool, error) {
 	var confirmed bool
@@ -21,3 +26,13 @@ func Confirm() (bool, error) {
 
 	return confirmed, nil
 }
+
+// IsStdinTerminal checks if the standard input is a terminal (TTY).
+func IsStdinTerminal() bool {
+	return term.IsTerminal(int(os.Stdin.Fd()))
+}
+
+// IsStdoutTerminal checks if the standard output is a terminal (TTY).
+func IsStdoutTerminal() bool {
+	return term.IsTerminal(int(os.Stdout.Fd()))
+}

internal/docker/client.go

@@ -11,7 +11,12 @@ import (
 	"github.com/docker/docker/client"
 )
 
+type Client struct {
+	*client.Client
+}
+
 // WaitDaemonReady waits for the Docker daemon to start and be ready to serve requests.
+// TODO(lhf): convert into a Client method.
 func WaitDaemonReady(ctx context.Context, cli *client.Client) error {
 	// Retry to ping the Docker daemon until it's ready or the context is canceled.
 	boff := backoff.WithContext(backoff.NewExponentialBackOff(

internal/docker/container.go

@@ -0,0 +1,70 @@
+package docker
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"time"
+
+	"github.com/containerd/errdefs"
+	"github.com/docker/docker/api/types/container"
+	"github.com/docker/docker/api/types/image"
+	"github.com/docker/go-connections/nat"
+)
+
+// CreateContainerWithImagePull creates a container. If the image is missing, it pulls the image first.
+func (cli *Client) CreateContainerWithImagePull(
+	ctx context.Context, name string, config *container.Config, hostConfig *container.HostConfig,
+) (container.CreateResponse, error) {
+	resp, err := cli.ContainerCreate(ctx, config, hostConfig, nil, nil, name)
+	if err == nil || !errdefs.IsNotFound(err) {
+		return resp, err
+	}
+
+	pullCh, err := cli.PullImage(ctx, config.Image, image.PullOptions{})
+	if err != nil {
+		return resp, fmt.Errorf("pull image: %w", err)
+	}
+
+	// Drain the pull channel until it's closed (image is fully pulled) or an error occurs.
+	// If the context is canceled during the pull, the channel will receive a context cancellation error.
+	for msg := range pullCh {
+		if msg.Err != nil {
+			return resp, fmt.Errorf("pull image: %w", msg.Err)
+		}
+	}
+
+	// Create container again after image pull.
+	if resp, err = cli.ContainerCreate(ctx, config, hostConfig, nil, nil, name); err != nil {
+		return resp, err
+	}
+
+	return resp, nil
+}
+
+// WaitPortPublished waits for a container port to be published on the host which happens asynchronously.
+func (cli *Client) WaitPortPublished(ctx context.Context, containerID string, port nat.Port) ([]nat.PortBinding, error) {
+	ctx, cancel := context.WithTimeout(ctx, 5*time.Second)
+	defer cancel()
+
+	for {
+		c, err := cli.ContainerInspect(ctx, containerID)
+		if err != nil {
+			return nil, fmt.Errorf("inspect container: %w", err)
+		}
+
+		binding, ok := c.NetworkSettings.Ports[port]
+		if ok && len(binding) > 0 {
+			return binding, nil
+		}
+
+		select {
+		case <-time.After(10 * time.Millisecond):
+		case <-ctx.Done():
+			if errors.Is(ctx.Err(), context.DeadlineExceeded) {
+				return nil, errors.New("timeout")
+			}
+			return nil, ctx.Err()
+		}
+	}
+}

internal/docker/image.go

@@ -0,0 +1,132 @@
+package docker
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"os"
+
+	dockercommand "github.com/docker/cli/cli/command"
+	dockerconfig "github.com/docker/cli/cli/config"
+	"github.com/docker/docker/api/types/image"
+	"github.com/docker/docker/api/types/registry"
+	"github.com/docker/docker/pkg/jsonmessage"
+)
+
+type PullPushImageMessage struct {
+	Message jsonmessage.JSONMessage
+	Err     error
+}
+
+// PullImage pulls an image and returns a channel to receive progress messages.
+func (cli *Client) PullImage(
+	ctx context.Context, image string, opts image.PullOptions,
+) (<-chan PullPushImageMessage, error) {
+	if opts.RegistryAuth == "" {
+		// Try to retrieve the authentication token for the image from the default local Docker config file.
+		if encodedAuth, err := RetrieveLocalDockerRegistryAuth(image); err == nil {
+			opts.RegistryAuth = encodedAuth
+		}
+	}
+
+	respBody, err := cli.ImagePull(ctx, image, opts)
+	if err != nil {
+		return nil, err
+	}
+
+	ch := make(chan PullPushImageMessage)
+	go processPullPushImageResp(ctx, respBody, ch)
+
+	return ch, nil
+}
+
+// PushImage pushes an image and returns a channel to receive progress messages.
+func (cli *Client) PushImage(
+	ctx context.Context, image string, opts image.PushOptions,
+) (<-chan PullPushImageMessage, error) {
+	if opts.RegistryAuth == "" {
+		// Try to retrieve the authentication token for the image from the default local Docker config file.
+		encodedAuth, _ := RetrieveLocalDockerRegistryAuth(image)
+		if encodedAuth == "" {
+			// If no credentials are found, provide an encoded empty auth config to work around the bug in Docker:
+			// https://github.com/moby/moby/issues/50729
+			var err error
+			if encodedAuth, err = registry.EncodeAuthConfig(registry.AuthConfig{}); err != nil {
+				return nil, fmt.Errorf("encode empty auth config: %w", err)
+			}
+		}
+		opts.RegistryAuth = encodedAuth
+	}
+
+	respBody, err := cli.ImagePush(ctx, image, opts)
+	if err != nil {
+		return nil, err
+	}
+
+	ch := make(chan PullPushImageMessage)
+	go processPullPushImageResp(ctx, respBody, ch)
+
+	return ch, nil
+}
+
+// processPullPushImageResp decodes JSON messages from the image pull/push response body and
+// sends them to the provided channel.
+func processPullPushImageResp(ctx context.Context, respBody io.ReadCloser, ch chan<- PullPushImageMessage) {
+	defer respBody.Close()
+	defer close(ch)
+
+	decoder := json.NewDecoder(respBody)
+	for {
+		var jm jsonmessage.JSONMessage
+		if err := decoder.Decode(&jm); err != nil {
+			if errors.Is(err, io.EOF) {
+				return
+			}
+			ch <- PullPushImageMessage{Err: fmt.Errorf("decode image pull/push message: %w", err)}
+			return
+		}
+
+		msg := PullPushImageMessage{Message: jm}
+		if jm.Error != nil {
+			msg.Err = errors.New(jm.Error.Message)
+		}
+
+		select {
+		case <-ctx.Done():
+			ch <- PullPushImageMessage{Err: ctx.Err()}
+			return
+		default:
+			ch <- msg
+		}
+	}
+}
+
+// RetrieveLocalDockerRegistryAuth retrieves the authentication token for the specified image from the local Docker
+// config file. It returns the encoded authentication token if it contains any credentials, or an empty string if
+// no credentials are found.
+func RetrieveLocalDockerRegistryAuth(image string) (string, error) {
+	// Try to retrieve the authentication token for the image from the default local Docker config file.
+	dockerConfig := dockerconfig.LoadDefaultConfigFile(os.Stderr)
+	encodedAuth, err := dockercommand.RetrieveAuthTokenFromImage(dockerConfig, image)
+	if err != nil {
+		return "", err
+	}
+	// The encodedAuth can be a base64-encoded "{}" (empty JSON object) or include a server address but no credentials.
+	// Return encodedAuth only if it contains any credentials.
+	auth, err := registry.DecodeAuthConfig(encodedAuth)
+	if err != nil {
+		return "", fmt.Errorf("decode auth config: %w", err)
+	}
+
+	if auth.Username == "" &&
+		auth.Password == "" &&
+		auth.Auth == "" &&
+		auth.IdentityToken == "" &&
+		auth.RegistryToken == "" {
+		return "", nil
+	}
+
+	return encodedAuth, nil
+}

internal/gitutil/state.go

@@ -0,0 +1,96 @@
+package gitutil
+
+import (
+	"fmt"
+	"os/exec"
+	"strconv"
+	"strings"
+	"time"
+)
+
+// GitState contains information about the Git repository state.
+type GitState struct {
+	// Date is the current commit datetime.
+	Date time.Time
+	// IsDirty indicates whether there are uncommitted changes.
+	IsDirty bool
+	// IsRepo indicates whether the directory is a Git repository.
+	IsRepo bool
+	// SHA is the full SHA-1 (40 characters) of the current commit.
+	SHA string
+}
+
+// InspectGitState inspects the Git repo state from the specified directory.
+// If the directory is not a Git repository, or if git utility is not available,
+// returns GitState with IsRepo=false and no error.
+func InspectGitState(dir string) (GitState, error) {
+	// Check if git command is available.
+	if _, err := exec.LookPath("git"); err != nil {
+		return GitState{}, nil
+	}
+
+	state := GitState{
+		IsRepo: isGitRepo(dir),
+	}
+	if !state.IsRepo {
+		return state, nil
+	}
+
+	// Get the current commit SHA.
+	sha, err := gitCommand(dir, "rev-parse", "HEAD")
+	if err != nil {
+		return state, fmt.Errorf("get current commit SHA: %w", err)
+	}
+	state.SHA = strings.TrimSpace(sha)
+
+	// Get the current commit timestamp.
+	timestamp, err := gitCommand(dir, "log", "-1", "--format=%ct")
+	if err != nil {
+		return state, fmt.Errorf("get current commit timestamp: %w", err)
+	}
+	seconds, err := strconv.ParseInt(strings.TrimSpace(timestamp), 10, 64)
+	if err != nil {
+		return state, fmt.Errorf("parse current commit timestamp: %w", err)
+	}
+	state.Date = time.Unix(seconds, 0).UTC()
+
+	// Check for uncommitted changes.
+	status, err := gitCommand(dir, "status", "--porcelain")
+	if err != nil {
+		return state, fmt.Errorf("check git status: %w", err)
+	}
+	state.IsDirty = strings.TrimSpace(status) != ""
+
+	return state, nil
+}
+
+// isGitRepo checks if the directory is a Git repository.
+func isGitRepo(dir string) bool {
+	_, err := gitCommand(dir, "rev-parse", "--git-dir")
+	return err == nil
+}
+
+// gitCommand runs a git command in the specified directory.
+func gitCommand(dir string, args ...string) (string, error) {
+	cmd := exec.Command("git", args...)
+	cmd.Dir = dir
+	output, err := cmd.Output()
+	if err != nil {
+		if exitErr, ok := err.(*exec.ExitError); ok {
+			return "", fmt.Errorf("git command failed: %s", exitErr.Stderr)
+		}
+		return "", err
+	}
+
+	return string(output), nil
+}
+
+// ShortSHA returns a truncated SHA with the specified length.
+// If length is negative, zero, or greater than the SHA length, it returns the full SHA.
+// Returns empty string if SHA is empty.
+func (s *GitState) ShortSHA(length int) string {
+	if length <= 0 || length > len(s.SHA) {
+		length = len(s.SHA)
+	}
+	return s.SHA[:length]
+}

internal/gitutil/state_test.go

@@ -0,0 +1,168 @@
+package gitutil
+
+import (
+	"os"
+	"os/exec"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestInspectGitState_GitNotAvailable(t *testing.T) {
+	// Reset the PATH env var to simulate git not being available.
+	t.Setenv("PATH", "")
+
+	tmpDir := t.TempDir()
+
+	state, err := InspectGitState(tmpDir)
+	require.NoError(t, err)
+
+	assert.False(t, state.IsRepo)
+	assert.Empty(t, state.SHA)
+	assert.Empty(t, state.ShortSHA(7))
+	assert.True(t, state.Date.IsZero())
+	assert.False(t, state.IsDirty)
+}
+
+func TestInspectGitState_NotARepo(t *testing.T) {
+	// Create a temporary directory that's not a git repo.
+	tmpDir := t.TempDir()
+
+	state, err := InspectGitState(tmpDir)
+	require.NoError(t, err)
+
+	// If tmpDir happens to be in a git repo (e.g., parent directory), skip this test.
+	// This is common when running tests from within the project itself.
+	if state.IsRepo {
+		t.Skip("tmpDir is within a git repository, skipping test")
+	}
+
+	assert.False(t, state.IsRepo)
+	assert.Empty(t, state.SHA)
+	assert.Empty(t, state.ShortSHA(7))
+	assert.True(t, state.Date.IsZero()) // Should have zero time if not a repo.
+	assert.False(t, state.IsDirty)
+}
+
+func TestInspectGitState_CleanRepo(t *testing.T) {
+	// Create a temporary git repo.
+	tmpDir := t.TempDir()
+	initGitRepo(t, tmpDir)
+
+	createAndCommitFile(t, tmpDir, "test.txt", "test content")
+
+	state, err := InspectGitState(tmpDir)
+	require.NoError(t, err)
+
+	assert.True(t, state.IsRepo)
+	assert.Len(t, state.SHA, 40)
+	assert.Equal(t, state.ShortSHA(7), state.SHA[:7])
+	assert.False(t, state.Date.IsZero())
+	assert.False(t, state.IsDirty)
+}
+
+func TestInspectGitState_DirtyRepo(t *testing.T) {
+	// Create a temporary git repo.
+	tmpDir := t.TempDir()
+	initGitRepo(t, tmpDir)
+
+	createAndCommitFile(t, tmpDir, "test.txt", "test content")
+
+	// Modify the file to make the repo dirty.
+	err := os.WriteFile(filepath.Join(tmpDir, "test.txt"), []byte("modified content"), 0o644)
+	require.NoError(t, err)
+
+	state, err := InspectGitState(tmpDir)
+	require.NoError(t, err)
+
+	assert.True(t, state.IsRepo)
+	assert.True(t, state.IsDirty)
+}
+
+func TestInspectGitState_UntrackedFiles(t *testing.T) {
+	// Create a temporary git repo.
+	tmpDir := t.TempDir()
+	initGitRepo(t, tmpDir)
+
+	createAndCommitFile(t, tmpDir, "test.txt", "test content")
+
+	// Add an untracked file.
+	err := os.WriteFile(filepath.Join(tmpDir, "untracked.txt"), []byte("untracked"), 0o644)
+	require.NoError(t, err)
+
+	state, err := InspectGitState(tmpDir)
+	require.NoError(t, err)
+
+	assert.True(t, state.IsRepo)
+	assert.True(t, state.IsDirty) // Untracked files make the repo dirty.
+}
+
+func TestGitState_ShortSHA(t *testing.T) {
+	state := &GitState{
+		SHA: "1234567890abcdef1234567890abcdef12345678",
+	}
+
+	tests := []struct {
+		name     string
+		length   int
+		expected string
+	}{
+		{"short 7", 7, "1234567"},
+		{"short 10", 10, "1234567890"},
+		{"full SHA", 40, "1234567890abcdef1234567890abcdef12345678"},
+		{"longer than SHA", 50, "1234567890abcdef1234567890abcdef12345678"},
+		{"negative", -42, "1234567890abcdef1234567890abcdef12345678"},
+		{"zero", 0, "1234567890abcdef1234567890abcdef12345678"},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := state.ShortSHA(tt.length)
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}
+
+func TestGitState_ShortSHA_Empty(t *testing.T) {
+	state := &GitState{
+		SHA: "",
+	}
+
+	result := state.ShortSHA(7)
+	assert.Empty(t, result)
+}
+
+// Helper functions.
+
+func initGitRepo(t *testing.T, dir string) {
+	t.Helper()
+
+	// Initialize git repo.
+	runGitCommand(t, dir, "init")
+	runGitCommand(t, dir, "config", "user.email", "test@example.com")
+	runGitCommand(t, dir, "config", "user.name", "Test User")
+}
+
+func createAndCommitFile(t *testing.T, dir, filename, content string) {
+	t.Helper()
+
+	// Create file.
+	path := filepath.Join(dir, filename)
+	err := os.WriteFile(path, []byte(content), 0o644)
+	require.NoError(t, err)
+
+	// Commit file.
+	runGitCommand(t, dir, "add", filename)
+	runGitCommand(t, dir, "commit", "-m", "Add "+filename)
+}
+
+func runGitCommand(t *testing.T, dir string, args ...string) {
+	t.Helper()
+
+	cmd := exec.Command("git", args...)
+	cmd.Dir = dir
+	output, err := cmd.CombinedOutput()
+	require.NoError(t, err, "git command failed: %s", output)
+}

internal/log/env.go

@@ -0,0 +1,19 @@
+package log
+
+import (
+	"log/slog"
+	"os"
+	"slices"
+	"strings"
+)
+
+func InitLoggerFromEnv() {
+	debugValues := []string{"1", "true", "yes"}
+	if slices.Contains(debugValues, strings.ToLower(os.Getenv("DEBUG"))) {
+		logger := slog.New(NewSlogTextHandler(os.Stderr, &slog.HandlerOptions{
+			Level: slog.LevelDebug,
+		}))
+		slog.SetDefault(logger)
+	}
+	slog.Debug("logger initialized")
+}

internal/machine/api/pb/docker.proto

@@ -5,6 +5,7 @@ package api;
 option go_package = "github.com/psviderski/uncloud/internal/machine/api/pb";
 
 import "google/protobuf/empty.proto";
+import "google/protobuf/timestamp.proto";
 import "internal/machine/api/pb/common.proto";
 
 service Docker {
@@ -15,11 +16,15 @@ service Docker {
   rpc ListContainers(ListContainersRequest) returns (ListContainersResponse);
   rpc RemoveContainer(RemoveContainerRequest) returns (google.protobuf.Empty);
 
+  rpc ExecContainer(stream ExecContainerRequest) returns (stream ExecContainerResponse);
+  rpc ContainerLogs(ContainerLogsRequest) returns (stream ContainerLogEntry);
+
   rpc PullImage(PullImageRequest) returns (stream JSONMessage);
   rpc InspectImage(InspectImageRequest) returns (InspectImageResponse);
   // InspectRemoteImage returns the image metadata for an image in a remote registry using the machine's
   // Docker auth credentials if necessary.
   rpc InspectRemoteImage(InspectRemoteImageRequest) returns (InspectRemoteImageResponse);
+  rpc ListImages(ListImagesRequest) returns (ListImagesResponse);
 
   rpc CreateVolume(CreateVolumeRequest) returns (CreateVolumeResponse);
   rpc ListVolumes(ListVolumesRequest) returns (ListVolumesResponse);
@@ -91,6 +96,64 @@ message RemoveContainerRequest {
   bytes options = 2;
 }
 
+message ExecContainerRequest {
+  oneof payload {
+    // Initial configuration for the exec session. Must be sent as the first message.
+    ExecConfig config = 1;
+    // Raw stdin data to be written to the exec process.
+    bytes stdin = 2;
+    // TTY resize event (only used when TTY is enabled).
+    ResizeEvent resize = 3;
+  }
+}
+
+message ExecConfig {
+  // Container ID to execute the command in.
+  string container_id = 1;
+  // JSON serialised ExecOptions
+  bytes options = 2;
+}
+
+message ResizeEvent {
+  uint32 height = 1;
+  uint32 width = 2;
+}
+
+message ExecContainerResponse {
+  oneof payload {
+    // Exec instance ID returned after creating the exec.
+    string exec_id = 1;
+    // Raw stdout data from the exec process.
+    bytes stdout = 2;
+    // Raw stderr data from the exec process (only when TTY is disabled).
+    bytes stderr = 3;
+    // Exit code of the exec process. Sent as the final message.
+    int32 exit_code = 4;
+  }
+}
+
+message ContainerLogsRequest {
+  string container_id = 1;
+  // Options for logs retrieval.
+  bool follow = 2;
+  int32 tail = 3; // -1 means all
+  string since = 4; // https://www.rfc-editor.org/rfc/rfc3339.html timestamp or Go duration string
+  string until = 5; // https://www.rfc-editor.org/rfc/rfc3339.html timestamp or Go duration string
+}
+
+message ContainerLogEntry {
+  enum StreamType {
+    UNKNOWN = 0;
+    STDOUT = 1;
+    STDERR = 2;
+    HEARTBEAT = 3;
+  }
+  StreamType stream = 1;
+  google.protobuf.Timestamp timestamp = 2;
+  // Log line content. Empty for heartbeat entries.
+  bytes message = 3;
+}
+
 message PullImageRequest {
   string image = 1;
   // JSON serialised image.PullOptions.
@@ -113,7 +176,7 @@ message InspectImageResponse {
 
 message Image {
   Metadata metadata = 1;
-  // JSON serialised types.ImageInspect.
+  // JSON serialised image.InspectResponse.
   bytes image = 2;
 }
 
@@ -134,6 +197,24 @@ message RemoteImage {
   bytes manifest = 3;
 }
 
+message ListImagesRequest {
+  // JSON serialised image.ListOptions.
+  bytes options = 1;
+}
+
+message ListImagesResponse {
+  // Must contain only one repeated messages field to allow broadcasting ListImages requests to multiple machines.
+  repeated MachineImages messages = 1;
+}
+
+message MachineImages {
+  Metadata metadata = 1;
+  // JSON serialised []image.Summary.
+  bytes images = 2;
+  // True if Docker uses the containerd image store, false if it uses its internal image store.
+  bool containerd_store = 3;
+}
+
 message CreateVolumeRequest {
   // JSON serialised volume.CreateOptions.
   bytes options = 1;

internal/machine/api/pb/docker_grpc.pb.go

@@ -26,9 +26,12 @@ const (
 	Docker_StopContainer_FullMethodName           = "/api.Docker/StopContainer"
 	Docker_ListContainers_FullMethodName          = "/api.Docker/ListContainers"
 	Docker_RemoveContainer_FullMethodName         = "/api.Docker/RemoveContainer"
+	Docker_ExecContainer_FullMethodName           = "/api.Docker/ExecContainer"
+	Docker_ContainerLogs_FullMethodName           = "/api.Docker/ContainerLogs"
 	Docker_PullImage_FullMethodName               = "/api.Docker/PullImage"
 	Docker_InspectImage_FullMethodName            = "/api.Docker/InspectImage"
 	Docker_InspectRemoteImage_FullMethodName      = "/api.Docker/InspectRemoteImage"
+	Docker_ListImages_FullMethodName              = "/api.Docker/ListImages"
 	Docker_CreateVolume_FullMethodName            = "/api.Docker/CreateVolume"
 	Docker_ListVolumes_FullMethodName             = "/api.Docker/ListVolumes"
 	Docker_RemoveVolume_FullMethodName            = "/api.Docker/RemoveVolume"
@@ -48,11 +51,14 @@ type DockerClient interface {
 	StopContainer(ctx context.Context, in *StopContainerRequest, opts ...grpc.CallOption) (*emptypb.Empty, error)
 	ListContainers(ctx context.Context, in *ListContainersRequest, opts ...grpc.CallOption) (*ListContainersResponse, error)
 	RemoveContainer(ctx context.Context, in *RemoveContainerRequest, opts ...grpc.CallOption) (*emptypb.Empty, error)
+	ExecContainer(ctx context.Context, opts ...grpc.CallOption) (grpc.BidiStreamingClient[ExecContainerRequest, ExecContainerResponse], error)
+	ContainerLogs(ctx context.Context, in *ContainerLogsRequest, opts ...grpc.CallOption) (grpc.ServerStreamingClient[ContainerLogEntry], error)
 	PullImage(ctx context.Context, in *PullImageRequest, opts ...grpc.CallOption) (grpc.ServerStreamingClient[JSONMessage], error)
 	InspectImage(ctx context.Context, in *InspectImageRequest, opts ...grpc.CallOption) (*InspectImageResponse, error)
 	// InspectRemoteImage returns the image metadata for an image in a remote registry using the machine's
 	// Docker auth credentials if necessary.
 	InspectRemoteImage(ctx context.Context, in *InspectRemoteImageRequest, opts ...grpc.CallOption) (*InspectRemoteImageResponse, error)
+	ListImages(ctx context.Context, in *ListImagesRequest, opts ...grpc.CallOption) (*ListImagesResponse, error)
 	CreateVolume(ctx context.Context, in *CreateVolumeRequest, opts ...grpc.CallOption) (*CreateVolumeResponse, error)
 	ListVolumes(ctx context.Context, in *ListVolumesRequest, opts ...grpc.CallOption) (*ListVolumesResponse, error)
 	RemoveVolume(ctx context.Context, in *RemoveVolumeRequest, opts ...grpc.CallOption) (*emptypb.Empty, error)
@@ -130,9 +136,41 @@ func (c *dockerClient) RemoveContainer(ctx context.Context, in *RemoveContainerR
 	return out, nil
 }
 
+func (c *dockerClient) ExecContainer(ctx context.Context, opts ...grpc.CallOption) (grpc.BidiStreamingClient[ExecContainerRequest, ExecContainerResponse], error) {
+	cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
+	stream, err := c.cc.NewStream(ctx, &Docker_ServiceDesc.Streams[0], Docker_ExecContainer_FullMethodName, cOpts...)
+	if err != nil {
+		return nil, err
+	}
+	x := &grpc.GenericClientStream[ExecContainerRequest, ExecContainerResponse]{ClientStream: stream}
+	return x, nil
+}
+
+// This type alias is provided for backwards compatibility with existing code that references the prior non-generic stream type by name.
+type Docker_ExecContainerClient = grpc.BidiStreamingClient[ExecContainerRequest, ExecContainerResponse]
+
+func (c *dockerClient) ContainerLogs(ctx context.Context, in *ContainerLogsRequest, opts ...grpc.CallOption) (grpc.ServerStreamingClient[ContainerLogEntry], error) {
+	cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
+	stream, err := c.cc.NewStream(ctx, &Docker_ServiceDesc.Streams[1], Docker_ContainerLogs_FullMethodName, cOpts...)
+	if err != nil {
+		return nil, err
+	}
+	x := &grpc.GenericClientStream[ContainerLogsRequest, ContainerLogEntry]{ClientStream: stream}
+	if err := x.ClientStream.SendMsg(in); err != nil {
+		return nil, err
+	}
+	if err := x.ClientStream.CloseSend(); err != nil {
+		return nil, err
+	}
+	return x, nil
+}
+
+// This type alias is provided for backwards compatibility with existing code that references the prior non-generic stream type by name.
+type Docker_ContainerLogsClient = grpc.ServerStreamingClient[ContainerLogEntry]
+
 func (c *dockerClient) PullImage(ctx context.Context, in *PullImageRequest, opts ...grpc.CallOption) (grpc.ServerStreamingClient[JSONMessage], error) {
 	cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
-	stream, err := c.cc.NewStream(ctx, &Docker_ServiceDesc.Streams[0], Docker_PullImage_FullMethodName, cOpts...)
+	stream, err := c.cc.NewStream(ctx, &Docker_ServiceDesc.Streams[2], Docker_PullImage_FullMethodName, cOpts...)
 	if err != nil {
 		return nil, err
 	}
@@ -169,6 +207,16 @@ func (c *dockerClient) InspectRemoteImage(ctx context.Context, in *InspectRemote
 	return out, nil
 }
 
+func (c *dockerClient) ListImages(ctx context.Context, in *ListImagesRequest, opts ...grpc.CallOption) (*ListImagesResponse, error) {
+	cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
+	out := new(ListImagesResponse)
+	err := c.cc.Invoke(ctx, Docker_ListImages_FullMethodName, in, out, cOpts...)
+	if err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
 func (c *dockerClient) CreateVolume(ctx context.Context, in *CreateVolumeRequest, opts ...grpc.CallOption) (*CreateVolumeResponse, error) {
 	cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
 	out := new(CreateVolumeResponse)
@@ -249,11 +297,14 @@ type DockerServer interface {
 	StopContainer(context.Context, *StopContainerRequest) (*emptypb.Empty, error)
 	ListContainers(context.Context, *ListContainersRequest) (*ListContainersResponse, error)
 	RemoveContainer(context.Context, *RemoveContainerRequest) (*emptypb.Empty, error)
+	ExecContainer(grpc.BidiStreamingServer[ExecContainerRequest, ExecContainerResponse]) error
+	ContainerLogs(*ContainerLogsRequest, grpc.ServerStreamingServer[ContainerLogEntry]) error
 	PullImage(*PullImageRequest, grpc.ServerStreamingServer[JSONMessage]) error
 	InspectImage(context.Context, *InspectImageRequest) (*InspectImageResponse, error)
 	// InspectRemoteImage returns the image metadata for an image in a remote registry using the machine's
 	// Docker auth credentials if necessary.
 	InspectRemoteImage(context.Context, *InspectRemoteImageRequest) (*InspectRemoteImageResponse, error)
+	ListImages(context.Context, *ListImagesRequest) (*ListImagesResponse, error)
 	CreateVolume(context.Context, *CreateVolumeRequest) (*CreateVolumeResponse, error)
 	ListVolumes(context.Context, *ListVolumesRequest) (*ListVolumesResponse, error)
 	RemoveVolume(context.Context, *RemoveVolumeRequest) (*emptypb.Empty, error)
@@ -289,6 +340,12 @@ func (UnimplementedDockerServer) ListContainers(context.Context, *ListContainers
 func (UnimplementedDockerServer) RemoveContainer(context.Context, *RemoveContainerRequest) (*emptypb.Empty, error) {
 	return nil, status.Errorf(codes.Unimplemented, "method RemoveContainer not implemented")
 }
+func (UnimplementedDockerServer) ExecContainer(grpc.BidiStreamingServer[ExecContainerRequest, ExecContainerResponse]) error {
+	return status.Errorf(codes.Unimplemented, "method ExecContainer not implemented")
+}
+func (UnimplementedDockerServer) ContainerLogs(*ContainerLogsRequest, grpc.ServerStreamingServer[ContainerLogEntry]) error {
+	return status.Errorf(codes.Unimplemented, "method ContainerLogs not implemented")
+}
 func (UnimplementedDockerServer) PullImage(*PullImageRequest, grpc.ServerStreamingServer[JSONMessage]) error {
 	return status.Errorf(codes.Unimplemented, "method PullImage not implemented")
 }
@@ -298,6 +355,9 @@ func (UnimplementedDockerServer) InspectImage(context.Context, *InspectImageRequ
 func (UnimplementedDockerServer) InspectRemoteImage(context.Context, *InspectRemoteImageRequest) (*InspectRemoteImageResponse, error) {
 	return nil, status.Errorf(codes.Unimplemented, "method InspectRemoteImage not implemented")
 }
+func (UnimplementedDockerServer) ListImages(context.Context, *ListImagesRequest) (*ListImagesResponse, error) {
+	return nil, status.Errorf(codes.Unimplemented, "method ListImages not implemented")
+}
 func (UnimplementedDockerServer) CreateVolume(context.Context, *CreateVolumeRequest) (*CreateVolumeResponse, error) {
 	return nil, status.Errorf(codes.Unimplemented, "method CreateVolume not implemented")
 }
@@ -448,6 +508,24 @@ func _Docker_RemoveContainer_Handler(srv interface{}, ctx context.Context, dec f
 	return interceptor(ctx, in, info, handler)
 }
 
+func _Docker_ExecContainer_Handler(srv interface{}, stream grpc.ServerStream) error {
+	return srv.(DockerServer).ExecContainer(&grpc.GenericServerStream[ExecContainerRequest, ExecContainerResponse]{ServerStream: stream})
+}
+
+// This type alias is provided for backwards compatibility with existing code that references the prior non-generic stream type by name.
+type Docker_ExecContainerServer = grpc.BidiStreamingServer[ExecContainerRequest, ExecContainerResponse]
+
+func _Docker_ContainerLogs_Handler(srv interface{}, stream grpc.ServerStream) error {
+	m := new(ContainerLogsRequest)
+	if err := stream.RecvMsg(m); err != nil {
+		return err
+	}
+	return srv.(DockerServer).ContainerLogs(m, &grpc.GenericServerStream[ContainerLogsRequest, ContainerLogEntry]{ServerStream: stream})
+}
+
+// This type alias is provided for backwards compatibility with existing code that references the prior non-generic stream type by name.
+type Docker_ContainerLogsServer = grpc.ServerStreamingServer[ContainerLogEntry]
+
 func _Docker_PullImage_Handler(srv interface{}, stream grpc.ServerStream) error {
 	m := new(PullImageRequest)
 	if err := stream.RecvMsg(m); err != nil {
@@ -495,6 +573,24 @@ func _Docker_InspectRemoteImage_Handler(srv interface{}, ctx context.Context, de
 	return interceptor(ctx, in, info, handler)
 }
 
+func _Docker_ListImages_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
+	in := new(ListImagesRequest)
+	if err := dec(in); err != nil {
+		return nil, err
+	}
+	if interceptor == nil {
+		return srv.(DockerServer).ListImages(ctx, in)
+	}
+	info := &grpc.UnaryServerInfo{
+		Server:     srv,
+		FullMethod: Docker_ListImages_FullMethodName,
+	}
+	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
+		return srv.(DockerServer).ListImages(ctx, req.(*ListImagesRequest))
+	}
+	return interceptor(ctx, in, info, handler)
+}
+
 func _Docker_CreateVolume_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
 	in := new(CreateVolumeRequest)
 	if err := dec(in); err != nil {
@@ -660,6 +756,10 @@ var Docker_ServiceDesc = grpc.ServiceDesc{
 			MethodName: "InspectRemoteImage",
 			Handler:    _Docker_InspectRemoteImage_Handler,
 		},
+		{
+			MethodName: "ListImages",
+			Handler:    _Docker_ListImages_Handler,
+		},
 		{
 			MethodName: "CreateVolume",
 			Handler:    _Docker_CreateVolume_Handler,
@@ -690,6 +790,17 @@ var Docker_ServiceDesc = grpc.ServiceDesc{
 		},
 	},
 	Streams: []grpc.StreamDesc{
+		{
+			StreamName:    "ExecContainer",
+			Handler:       _Docker_ExecContainer_Handler,
+			ServerStreams: true,
+			ClientStreams: true,
+		},
+		{
+			StreamName:    "ContainerLogs",
+			Handler:       _Docker_ContainerLogs_Handler,
+			ServerStreams: true,
+		},
 		{
 			StreamName:    "PullImage",
 			Handler:       _Docker_PullImage_Handler,

internal/machine/caddyconfig/caddyfile_test.go

@@ -7,7 +7,6 @@ import (
 	"testing"
 	"time"
 
-	"github.com/docker/docker/api/types"
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/network"
 	"github.com/psviderski/uncloud/internal/machine/docker"
@@ -848,15 +847,15 @@ func newContainerRecordWithCaddyConfig(serviceName, ip, caddyConfig, machineID s
 	return store.ContainerRecord{
 		Container: api.ServiceContainer{
 			Container: api.Container{
-				ContainerJSON: types.ContainerJSON{
-					ContainerJSONBase: &types.ContainerJSONBase{
+				InspectResponse: container.InspectResponse{
+					ContainerJSONBase: &container.ContainerJSONBase{
 						ID: serviceName + "-" + ip, // Add ID for stable sorting
-						State: &types.ContainerState{
+						State: &container.State{
 							Running: true,
 						},
 						Created: created.UTC().Format(time.RFC3339Nano),
 					},
-					NetworkSettings: &types.NetworkSettings{
+					NetworkSettings: &container.NetworkSettings{
 						Networks: map[string]*network.EndpointSettings{
 							docker.NetworkName: {
 								IPAddress: ip,
@@ -975,15 +974,15 @@ func newContainerRecordWithPorts(serviceName, ip string, ports []string, machine
 	return store.ContainerRecord{
 		Container: api.ServiceContainer{
 			Container: api.Container{
-				ContainerJSON: types.ContainerJSON{
-					ContainerJSONBase: &types.ContainerJSONBase{
+				InspectResponse: container.InspectResponse{
+					ContainerJSONBase: &container.ContainerJSONBase{
 						ID: serviceName + "-" + ip, // Add ID for stable sorting
-						State: &types.ContainerState{
+						State: &container.State{
 							Running: true,
 						},
 						Created: time.Now().UTC().Format(time.RFC3339Nano),
 					},
-					NetworkSettings: &types.NetworkSettings{
+					NetworkSettings: &container.NetworkSettings{
 						Networks: map[string]*network.EndpointSettings{
 							docker.NetworkName: {
 								IPAddress: ip,

internal/machine/caddyconfig/jsonconfig_test.go

@@ -4,7 +4,6 @@ import (
 	"strings"
 	"testing"
 
-	"github.com/docker/docker/api/types"
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/network"
 	"github.com/psviderski/uncloud/internal/machine/docker"
@@ -334,13 +333,13 @@ func TestGenerateJSONConfig(t *testing.T) {
 
 func newContainer(ip string, ports ...string) api.ServiceContainer {
 	portsLabel := strings.Join(ports, ",")
-	return api.ServiceContainer{Container: api.Container{ContainerJSON: types.ContainerJSON{
-		ContainerJSONBase: &types.ContainerJSONBase{
-			State: &types.ContainerState{
+	return api.ServiceContainer{Container: api.Container{InspectResponse: container.InspectResponse{
+		ContainerJSONBase: &container.ContainerJSONBase{
+			State: &container.State{
 				Running: true,
 			},
 		},
-		NetworkSettings: &types.NetworkSettings{
+		NetworkSettings: &container.NetworkSettings{
 			Networks: map[string]*network.EndpointSettings{
 				docker.NetworkName: {
 					IPAddress: ip,
@@ -357,13 +356,13 @@ func newContainer(ip string, ports ...string) api.ServiceContainer {
 
 func newContainerWithoutNetwork(ports ...string) api.ServiceContainer {
 	portsLabel := strings.Join(ports, ",")
-	return api.ServiceContainer{Container: api.Container{ContainerJSON: types.ContainerJSON{
-		ContainerJSONBase: &types.ContainerJSONBase{
-			State: &types.ContainerState{
+	return api.ServiceContainer{Container: api.Container{InspectResponse: container.InspectResponse{
+		ContainerJSONBase: &container.ContainerJSONBase{
+			State: &container.State{
 				Running: true,
 			},
 		},
-		NetworkSettings: &types.NetworkSettings{
+		NetworkSettings: &container.NetworkSettings{
 			Networks: map[string]*network.EndpointSettings{
 				"other-network": {
 					IPAddress: "172.17.0.2",

internal/machine/caddyconfig/server.go

@@ -27,9 +27,9 @@ func (s *Server) GetConfig(ctx context.Context, _ *emptypb.Empty) (*pb.GetCaddyC
 	caddyfile, modifiedAt, err := s.service.Caddyfile()
 	if err != nil {
 		if os.IsNotExist(err) {
-			return nil, status.Errorf(codes.NotFound, err.Error())
+			return nil, status.Error(codes.NotFound, err.Error())
 		}
-		return nil, status.Errorf(codes.Internal, err.Error())
+		return nil, status.Error(codes.Internal, err.Error())
 	}
 
 	return &pb.GetCaddyConfigResponse{

internal/machine/cluster.go

@@ -21,6 +21,7 @@ import (
 	"github.com/psviderski/uncloud/internal/machine/firewall"
 	"github.com/psviderski/uncloud/internal/machine/network"
 	"github.com/psviderski/uncloud/internal/machine/store"
+	"github.com/psviderski/unregistry"
 	"golang.org/x/sync/errgroup"
 	"google.golang.org/grpc"
 )
@@ -45,6 +46,8 @@ type clusterController struct {
 	// dnsServer is the embedded internal DNS server for the cluster listening on the machine IP.
 	dnsServer   *dns.Server
 	dnsResolver *dns.ClusterResolver
+	// unregistry is the embedded container registry that uses the local Docker (containerd) image store as its backend.
+	unregistry *unregistry.Registry
 
 	// stopped is a channel that is closed when the controller is stopped.
 	stopped chan struct{}
@@ -60,6 +63,7 @@ func newClusterController(
 	caddyfileCtrl *caddyconfig.Controller,
 	dnsServer *dns.Server,
 	dnsResolver *dns.ClusterResolver,
+	unregistry *unregistry.Registry,
 ) (*clusterController, error) {
 	slog.Info("Starting WireGuard network.")
 	wgnet, err := network.NewWireGuardNetwork()
@@ -80,6 +84,7 @@ func newClusterController(
 		caddyconfigCtrl: caddyfileCtrl,
 		dnsServer:       dnsServer,
 		dnsResolver:     dnsResolver,
+		unregistry:      unregistry,
 		stopped:         make(chan struct{}),
 	}, nil
 }
@@ -87,7 +92,7 @@ func newClusterController(
 func (cc *clusterController) Run(ctx context.Context) error {
 	defer close(cc.stopped)
 
-	if err := firewall.ConfigureIptablesChains(); err != nil {
+	if err := firewall.ConfigureIptablesChains(network.MachineIP(cc.state.Network.Subnet)); err != nil {
 		return fmt.Errorf("configure iptables chains: %w", err)
 	}
 
@@ -164,7 +169,7 @@ func (cc *clusterController) Run(ctx context.Context) error {
 		return nil
 	})
 
-	// Watch for endpoint changes and update the machine state accordingly.
+	// Watch for WireGuard peer endpoint changes and update the machine state accordingly.
 	errGroup.Go(func() error {
 		for {
 			select {
@@ -210,6 +215,16 @@ func (cc *clusterController) Run(ctx context.Context) error {
 		return nil
 	})
 
+	if cc.unregistry != nil {
+		errGroup.Go(func() error {
+			slog.Info("Starting unregistry server.")
+			if err := cc.unregistry.ListenAndServe(); err != nil {
+				return fmt.Errorf("unregistry server failed: %w", err)
+			}
+			return nil
+		})
+	}
+
 	// Wait for the context to be done and stop the network API server.
 	<-ctx.Done()
 	slog.Info("Stopping network API server.")
@@ -217,6 +232,19 @@ func (cc *clusterController) Run(ctx context.Context) error {
 	cc.server.GracefulStop()
 	slog.Info("Network API server stopped.")
 
+	// Stop the unregistry server with a timeout if it was started.
+	if cc.unregistry != nil {
+		unregTimeout := 30 * time.Second
+		slog.Info("Stopping unregistry server.", "timeout", unregTimeout)
+		unregCtx, cancel := context.WithTimeout(context.Background(), unregTimeout)
+		defer cancel()
+
+		if err = cc.unregistry.Shutdown(unregCtx); err != nil {
+			return fmt.Errorf("unregistry server forced to shutdown: %w", err)
+		}
+		slog.Info("Unregistry server stopped.")
+	}
+
 	// Wait for all controllers to finish.
 	err = errGroup.Wait()
 
@@ -330,6 +358,14 @@ func (cc *clusterController) handleMachineChanges(ctx context.Context) error {
 					slog.Error("Failed to list machines.", "err", err)
 					continue
 				}
+				// Skip reconfiguration if the machines list is empty. This can happen when joining the cluster.
+				// Corrosion can notifies about table changes before the data is fully replicated.
+				// Reconfiguring with an empty list would remove all peers and lock this machine out of the cluster.
+				// See https://github.com/psviderski/uncloud/issues/155.
+				if len(machines) == 0 {
+					slog.Debug("Skipping peer reconfiguration: machines list in store is empty.")
+					continue
+				}
 				if err = cc.configurePeers(machines); err != nil {
 					slog.Error("Failed to configure peers.", "err", err)
 				}

internal/machine/cluster/dns.go

@@ -44,7 +44,7 @@ func (c *Cluster) ReserveDomain(ctx context.Context, req *pb.ReserveDomainReques
 	dnsClient := dns.NewClient()
 	name, token, err := dnsClient.ReserveDomain(req.Endpoint)
 	if err != nil {
-		return nil, status.Errorf(codes.Internal, err.Error())
+		return nil, status.Error(codes.Internal, err.Error())
 	}
 
 	domain := uncloudDNSDomain{

internal/machine/constants/constants.go

@@ -3,4 +3,6 @@ package constants
 const (
 	// MachineAPIPort is the port for the Machine API service on the management WireGuard network.
 	MachineAPIPort = 51000
+	// UnregistryPort is the port for the embedded container registry listening on the machine IP.
+	UnregistryPort = 5000
 )

internal/machine/corroservice/docker.go

@@ -8,6 +8,7 @@ import (
 	"path/filepath"
 	"time"
 
+	"github.com/containerd/errdefs"
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/image"
 	"github.com/docker/docker/api/types/mount"
@@ -30,7 +31,7 @@ type DockerService struct {
 func (s *DockerService) Start(ctx context.Context) error {
 	_, err := s.Client.ContainerInspect(ctx, s.Name)
 	if err != nil {
-		if !client.IsErrNotFound(err) {
+		if !errdefs.IsNotFound(err) {
 			return fmt.Errorf("inspect container %q: %w", s.Name, err)
 		}
 		if err = s.startNewContainer(ctx); err != nil {
@@ -114,7 +115,7 @@ func (s *DockerService) hostConfig() *container.HostConfig {
 func (s *DockerService) startNewContainer(ctx context.Context) error {
 	_, err := s.Client.ContainerCreate(ctx, s.containerConfig(), s.hostConfig(), nil, nil, s.Name)
 	if err != nil {
-		if !client.IsErrNotFound(err) {
+		if !errdefs.IsNotFound(err) {
 			return fmt.Errorf("create container: %w", err)
 		}
 

internal/machine/dns/resolver.go

@@ -91,6 +91,11 @@ func (r *ClusterResolver) updateServiceIPs(containers []store.ContainerRecord) {
 		newServiceIPs[ctr.ServiceName()] = append(newServiceIPs[ctr.ServiceName()], ip)
 		// Also add the service ID as a valid lookup.
 		newServiceIPs[ctr.ServiceID()] = append(newServiceIPs[ctr.ServiceID()], ip)
+
+		// Add <machine-id>.m.<service-name> as a lookup
+		serviceNameWithMachineID := record.MachineID + ".m." + ctr.ServiceName()
+		newServiceIPs[serviceNameWithMachineID] = append(newServiceIPs[serviceNameWithMachineID], ip)
+
 		containersCount++
 	}
 
@@ -99,7 +104,7 @@ func (r *ClusterResolver) updateServiceIPs(containers []store.ContainerRecord) {
 	r.serviceIPs = newServiceIPs
 	r.mu.Unlock()
 
-	r.log.Debug("DNS records updated.", "services", len(newServiceIPs)/2, "containers", containersCount)
+	r.log.Debug("DNS records updated.", "services", len(newServiceIPs)/3, "containers", containersCount)
 }
 
 // Resolve returns IP addresses of the service containers.

internal/machine/dns/server.go

@@ -8,6 +8,7 @@ import (
 	"math/rand/v2"
 	"net"
 	"net/netip"
+	"slices"
 	"strconv"
 	"strings"
 	"sync"
@@ -40,6 +41,7 @@ type Resolver interface {
 // to upstream DNS servers.
 type Server struct {
 	listenAddr      netip.Addr
+	localSubnet     netip.Prefix
 	resolver        Resolver
 	upstreamServers []netip.AddrPort
 
@@ -53,7 +55,7 @@ type Server struct {
 // NewServer creates a new DNS server with the given configuration.
 // If upstreams is nil, nameservers from /etc/resolv.conf will be used. An empty upstreams list means to only resolve
 // internal DNS queries and not forward any external queries.
-func NewServer(listenAddr netip.Addr, resolver Resolver, upstreams []netip.AddrPort) (*Server, error) {
+func NewServer(listenAddr netip.Addr, localSubnet netip.Prefix, resolver Resolver, upstreams []netip.AddrPort) (*Server, error) {
 	if !listenAddr.IsValid() {
 		return nil, fmt.Errorf("invalid listen address: %s", listenAddr)
 	}
@@ -87,6 +89,7 @@ func NewServer(listenAddr netip.Addr, resolver Resolver, upstreams []netip.AddrP
 
 	return &Server{
 		listenAddr:       listenAddr,
+		localSubnet:      localSubnet,
 		resolver:         resolver,
 		upstreamServers:  upstreams,
 		forwardSemaphore: make(chan struct{}, maxConcurrentForwards),
@@ -287,7 +290,7 @@ func (s *Server) forwardRequest(req *dns.Msg, proto string) (*dns.Msg, error) {
 // handleAQuery processes an A query for the internal domain and returns A records for the requested name.
 // The internal domain suffix is already stripped from the name. An empty list is returned if no records are found.
 func (s *Server) handleAQuery(name string) []dns.RR {
-	serviceName := trimInternalDomain(name)
+	serviceName, mode := extractModeFromDomain(trimInternalDomain(name))
 	ips := s.resolver.Resolve(serviceName)
 	if len(ips) == 0 {
 		s.log.Debug("Failed to resolve service name.", "service", serviceName)
@@ -296,10 +299,28 @@ func (s *Server) handleAQuery(name string) []dns.RR {
 	s.log.Debug("Resolved service name.", "service", serviceName, "ips", ips)
 
 	if len(ips) > 1 {
-		// TODO: sort by proximity to the requesting container/machine. For now, just shuffle the IPs.
+		// Shuffle the IPs to approximate round-robin.
+		// We want to do this as a baseline for "nearest" mode, as well.
 		rand.Shuffle(len(ips), func(i, j int) {
 			ips[i], ips[j] = ips[j], ips[i]
 		})
+
+		// Default (mode == "") currently behaves the same as round-robin,
+		// and nothing additional to do for round-robin (mode == "rr").
+
+		if mode == "nearest" {
+			// Sort IPs on local subnet to the top.
+			slices.SortFunc(ips, func(a, b netip.Addr) int {
+				aIsLocal := s.localSubnet.Contains(a)
+				bIsLocal := s.localSubnet.Contains(b)
+				if aIsLocal && !bIsLocal {
+					return -1
+				} else if bIsLocal && !aIsLocal {
+					return 1
+				}
+				return 0
+			})
+		}
 	}
 
 	// Create A records for each IP.
@@ -345,3 +366,13 @@ func trimInternalDomain(name string) string {
 
 	return strings.TrimSuffix(name, "."+InternalDomain)
 }
+
+func extractModeFromDomain(name string) (string, string) {
+	modes := []string{"nearest", "rr"}
+	for _, mode := range modes {
+		if cut, found := strings.CutPrefix(name, mode+"."); found {
+			return cut, mode
+		}
+	}
+	return name, ""
+}

internal/machine/docker/client.go

@@ -8,7 +8,6 @@ import (
 	"io"
 
 	"github.com/distribution/reference"
-	"github.com/docker/docker/api/types"
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/network"
 	"github.com/docker/docker/api/types/volume"
@@ -16,6 +15,7 @@ import (
 	"github.com/docker/docker/pkg/jsonmessage"
 	regtypes "github.com/google/go-containerregistry/pkg/v1/types"
 	ocispec "github.com/opencontainers/image-spec/specs-go/v1"
+	"github.com/psviderski/uncloud/internal/docker"
 	"github.com/psviderski/uncloud/internal/machine/api/pb"
 	"github.com/psviderski/uncloud/pkg/api"
 	"google.golang.org/grpc"
@@ -24,16 +24,18 @@ import (
 )
 
 // Client is a gRPC client for the Docker service that provides a similar interface to the Docker HTTP client.
+// TODO: it doesn't seem there is much value in having this intermediate Docker client.
+// Consider merging it into the main pkg/client.
 type Client struct {
 	conn       *grpc.ClientConn
-	grpcClient pb.DockerClient
+	GRPCClient pb.DockerClient
 }
 
 // NewClient creates a new Docker gRPC client with the provided gRPC connection.
 func NewClient(conn *grpc.ClientConn) *Client {
 	return &Client{
 		conn:       conn,
-		grpcClient: pb.NewDockerClient(conn),
+		GRPCClient: pb.NewDockerClient(conn),
 	}
 }
 
@@ -70,7 +72,7 @@ func (c *Client) CreateContainer(
 		return resp, fmt.Errorf("marshal platform: %w", err)
 	}
 
-	grpcResp, err := c.grpcClient.CreateContainer(ctx, &pb.CreateContainerRequest{
+	grpcResp, err := c.GRPCClient.CreateContainer(ctx, &pb.CreateContainerRequest{
 		Config:        configBytes,
 		HostConfig:    hostConfigBytes,
 		NetworkConfig: networkingConfigBytes,
@@ -91,10 +93,10 @@ func (c *Client) CreateContainer(
 }
 
 // InspectContainer returns the container information for the given container ID.
-func (c *Client) InspectContainer(ctx context.Context, id string) (types.ContainerJSON, error) {
-	var resp types.ContainerJSON
+func (c *Client) InspectContainer(ctx context.Context, id string) (container.InspectResponse, error) {
+	var resp container.InspectResponse
 
-	grpcResp, err := c.grpcClient.InspectContainer(ctx, &pb.InspectContainerRequest{Id: id})
+	grpcResp, err := c.GRPCClient.InspectContainer(ctx, &pb.InspectContainerRequest{Id: id})
 	if err != nil {
 		if status.Convert(err).Code() == codes.NotFound {
 			return resp, errdefs.NotFound(err)
@@ -115,7 +117,7 @@ func (c *Client) StartContainer(ctx context.Context, id string, opts container.S
 		return fmt.Errorf("marshal options: %w", err)
 	}
 
-	_, err = c.grpcClient.StartContainer(ctx, &pb.StartContainerRequest{
+	_, err = c.GRPCClient.StartContainer(ctx, &pb.StartContainerRequest{
 		Id:      id,
 		Options: optsBytes,
 	})
@@ -134,7 +136,7 @@ func (c *Client) StopContainer(ctx context.Context, id string, opts container.St
 		return fmt.Errorf("marshal options: %w", err)
 	}
 
-	_, err = c.grpcClient.StopContainer(ctx, &pb.StopContainerRequest{
+	_, err = c.GRPCClient.StopContainer(ctx, &pb.StopContainerRequest{
 		Id:      id,
 		Options: optsBytes,
 	})
@@ -148,7 +150,7 @@ func (c *Client) StopContainer(ctx context.Context, id string, opts container.St
 
 type MachineContainers struct {
 	Metadata   *pb.Metadata
-	Containers []types.ContainerJSON
+	Containers []container.InspectResponse
 }
 
 func (c *Client) ListContainers(ctx context.Context, opts container.ListOptions) ([]MachineContainers, error) {
@@ -157,7 +159,7 @@ func (c *Client) ListContainers(ctx context.Context, opts container.ListOptions)
 		return nil, fmt.Errorf("marshal options: %w", err)
 	}
 
-	resp, err := c.grpcClient.ListContainers(ctx, &pb.ListContainersRequest{Options: optsBytes})
+	resp, err := c.GRPCClient.ListContainers(ctx, &pb.ListContainersRequest{Options: optsBytes})
 	if err != nil {
 		return nil, err
 	}
@@ -184,7 +186,7 @@ func (c *Client) RemoveContainer(ctx context.Context, id string, opts container.
 		return fmt.Errorf("marshal options: %w", err)
 	}
 
-	_, err = c.grpcClient.RemoveContainer(ctx, &pb.RemoveContainerRequest{
+	_, err = c.GRPCClient.RemoveContainer(ctx, &pb.RemoveContainerRequest{
 		Id:      id,
 		Options: optsBytes,
 	})
@@ -205,23 +207,18 @@ type PullOptions struct {
 	Platform     string
 }
 
-type PullImageMessage struct {
-	Message jsonmessage.JSONMessage
-	Err     error
-}
-
-func (c *Client) PullImage(ctx context.Context, image string, opts PullOptions) (<-chan PullImageMessage, error) {
+func (c *Client) PullImage(ctx context.Context, image string, opts PullOptions) (<-chan docker.PullPushImageMessage, error) {
 	optsBytes, err := json.Marshal(opts)
 	if err != nil {
 		return nil, fmt.Errorf("marshal options: %w", err)
 	}
 
-	stream, err := c.grpcClient.PullImage(ctx, &pb.PullImageRequest{Image: image, Options: optsBytes})
+	stream, err := c.GRPCClient.PullImage(ctx, &pb.PullImageRequest{Image: image, Options: optsBytes})
 	if err != nil {
 		return nil, err
 	}
 
-	ch := make(chan PullImageMessage)
+	ch := make(chan docker.PullPushImageMessage)
 
 	go func() {
 		defer close(ch)
@@ -232,16 +229,16 @@ func (c *Client) PullImage(ctx context.Context, image string, opts PullOptions)
 				return
 			}
 			if err != nil {
-				ch <- PullImageMessage{Err: err}
+				ch <- docker.PullPushImageMessage{Err: err}
 				return
 			}
 
 			var jm jsonmessage.JSONMessage
 			if err = json.Unmarshal(msg.Message, &jm); err != nil {
-				ch <- PullImageMessage{Err: fmt.Errorf("unmarshal JSON message: %w", err)}
+				ch <- docker.PullPushImageMessage{Err: fmt.Errorf("unmarshal JSON message: %w", err)}
 				return
 			}
-			ch <- PullImageMessage{Message: jm}
+			ch <- docker.PullPushImageMessage{Message: jm}
 		}
 	}()
 
@@ -250,7 +247,7 @@ func (c *Client) PullImage(ctx context.Context, image string, opts PullOptions)
 
 // InspectImage returns the image information for the given image ID. The request may be sent to multiple machines.
 func (c *Client) InspectImage(ctx context.Context, id string) ([]api.MachineImage, error) {
-	resp, err := c.grpcClient.InspectImage(ctx, &pb.InspectImageRequest{Id: id})
+	resp, err := c.GRPCClient.InspectImage(ctx, &pb.InspectImageRequest{Id: id})
 	if err != nil {
 		// If the request was sent to only one machine, err is an actual error from the machine.
 		if status.Convert(err).Code() == codes.NotFound {
@@ -288,7 +285,7 @@ func (c *Client) InspectImage(ctx context.Context, id string) ([]api.MachineImag
 // credentials if necessary. If the response from a machine doesn't contain an error, the api.RemoteImage will either
 // contain an IndexManifest or an ImageManifest.
 func (c *Client) InspectRemoteImage(ctx context.Context, id string) ([]api.MachineRemoteImage, error) {
-	resp, err := c.grpcClient.InspectRemoteImage(ctx, &pb.InspectRemoteImageRequest{Id: id})
+	resp, err := c.GRPCClient.InspectRemoteImage(ctx, &pb.InspectRemoteImageRequest{Id: id})
 	if err != nil {
 		return nil, err
 	}
@@ -365,7 +362,7 @@ func (c *Client) CreateVolume(ctx context.Context, opts volume.CreateOptions) (v
 		return vol, fmt.Errorf("marshal options: %w", err)
 	}
 
-	resp, err := c.grpcClient.CreateVolume(ctx, &pb.CreateVolumeRequest{Options: optsBytes})
+	resp, err := c.GRPCClient.CreateVolume(ctx, &pb.CreateVolumeRequest{Options: optsBytes})
 	if err != nil {
 		return vol, err
 	}
@@ -377,20 +374,18 @@ func (c *Client) CreateVolume(ctx context.Context, opts volume.CreateOptions) (v
 	return vol, nil
 }
 
-// MachineVolumes represents a volume list response from a machine.
 type MachineVolumes struct {
 	Metadata *pb.Metadata
 	Response volume.ListResponse
 }
 
-// ListVolumes returns a list of all volumes matching the filter.
 func (c *Client) ListVolumes(ctx context.Context, opts volume.ListOptions) ([]MachineVolumes, error) {
 	optsBytes, err := json.Marshal(opts)
 	if err != nil {
 		return nil, fmt.Errorf("marshal options: %w", err)
 	}
 
-	resp, err := c.grpcClient.ListVolumes(ctx, &pb.ListVolumesRequest{Options: optsBytes})
+	resp, err := c.GRPCClient.ListVolumes(ctx, &pb.ListVolumesRequest{Options: optsBytes})
 	if err != nil {
 		return nil, err
 	}
@@ -412,7 +407,7 @@ func (c *Client) ListVolumes(ctx context.Context, opts volume.ListOptions) ([]Ma
 
 // RemoveVolume removes a volume with the given ID.
 func (c *Client) RemoveVolume(ctx context.Context, id string, force bool) error {
-	_, err := c.grpcClient.RemoveVolume(ctx, &pb.RemoveVolumeRequest{
+	_, err := c.GRPCClient.RemoveVolume(ctx, &pb.RemoveVolumeRequest{
 		Id:    id,
 		Force: force,
 	})
@@ -435,7 +430,7 @@ func (c *Client) CreateServiceContainer(
 	if err != nil {
 		return resp, fmt.Errorf("marshal service spec: %w", err)
 	}
-	grpcResp, err := c.grpcClient.CreateServiceContainer(ctx, &pb.CreateServiceContainerRequest{
+	grpcResp, err := c.GRPCClient.CreateServiceContainer(ctx, &pb.CreateServiceContainerRequest{
 		ServiceId:     serviceID,
 		ServiceSpec:   specBytes,
 		ContainerName: containerName,
@@ -458,7 +453,7 @@ func (c *Client) CreateServiceContainer(
 func (c *Client) InspectServiceContainer(ctx context.Context, id string) (api.ServiceContainer, error) {
 	var resp api.ServiceContainer
 
-	grpcResp, err := c.grpcClient.InspectServiceContainer(ctx, &pb.InspectContainerRequest{Id: id})
+	grpcResp, err := c.GRPCClient.InspectServiceContainer(ctx, &pb.InspectContainerRequest{Id: id})
 	if err != nil {
 		if status.Convert(err).Code() == codes.NotFound {
 			return resp, errdefs.NotFound(err)
@@ -491,7 +486,7 @@ func (c *Client) ListServiceContainers(
 		return nil, fmt.Errorf("marshal options: %w", err)
 	}
 
-	resp, err := c.grpcClient.ListServiceContainers(ctx, &pb.ListServiceContainersRequest{
+	resp, err := c.GRPCClient.ListServiceContainers(ctx, &pb.ListServiceContainersRequest{
 		ServiceId: serviceNameOrID,
 		Options:   optsBytes,
 	})
@@ -530,7 +525,7 @@ func (c *Client) RemoveServiceContainer(ctx context.Context, id string, opts con
 		return fmt.Errorf("marshal options: %w", err)
 	}
 
-	_, err = c.grpcClient.RemoveServiceContainer(ctx, &pb.RemoveContainerRequest{
+	_, err = c.GRPCClient.RemoveServiceContainer(ctx, &pb.RemoveContainerRequest{
 		Id:      id,
 		Options: optsBytes,
 	})

internal/machine/docker/client_exec.go

@@ -0,0 +1,303 @@
+package docker
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"log/slog"
+	"os"
+	"os/signal"
+
+	"github.com/moby/term"
+	"github.com/psviderski/uncloud/internal/machine/api/pb"
+	"github.com/psviderski/uncloud/pkg/api"
+	"golang.org/x/sync/errgroup"
+	"golang.org/x/sys/unix"
+)
+
+// ExecConfig contains options for executing a command in a container.
+type ExecConfig struct {
+	// Container ID to execute the command in.
+	ContainerID string
+	// Exec configuration.
+	Options api.ExecOptions
+}
+
+// sendResizeRequest sends a terminal resize request to the exec stream.
+func sendResizeRequest(stream pb.Docker_ExecContainerClient, size *term.Winsize) error {
+	slog.Debug("sending resize request", "width", size.Width, "height", size.Height)
+	return stream.Send(
+		&pb.ExecContainerRequest{
+			Payload: &pb.ExecContainerRequest_Resize{
+				Resize: &pb.ResizeEvent{
+					Height: uint32(size.Height),
+					Width:  uint32(size.Width),
+				},
+			},
+		})
+}
+
+// setupTerminal configures the terminal for interactive TTY sessions.
+// It checks if stdin is a terminal, sets it to raw mode, and sets up resize handling.
+// Returns a cleanup function to restore terminal state, or an error.
+func setupTerminal(ctx context.Context, stream pb.Docker_ExecContainerClient) (func(), error) {
+	inFd, isTerminal := term.GetFdInfo(os.Stdin)
+	if !isTerminal {
+		return nil, fmt.Errorf("stdin is not a terminal")
+	}
+
+	// Set terminal to raw mode
+	oldState, err := term.SetRawTerminal(inFd)
+	if err != nil {
+		return nil, fmt.Errorf("set raw terminal: %w", err)
+	}
+
+	// Cleanup function
+	restoreFunc := func() {
+		_ = term.RestoreTerminal(inFd, oldState)
+	}
+
+	// Set up resize handling
+	if err := handleTerminalResize(ctx, inFd, stream); err != nil {
+		restoreFunc()
+		return nil, err
+	}
+
+	return restoreFunc, nil
+}
+
+// handleTerminalResize sends initial window size and handles window resize signals for TTY sessions.
+func handleTerminalResize(ctx context.Context, inFd uintptr, stream pb.Docker_ExecContainerClient) error {
+	// Handle window resize signals
+	sigCh := make(chan os.Signal, 1)
+	signal.Notify(sigCh, unix.SIGWINCH)
+
+	// Send initial window size
+	if size, err := term.GetWinsize(inFd); err == nil {
+		_ = sendResizeRequest(stream, size)
+	}
+
+	go func() {
+		defer signal.Stop(sigCh)
+		for {
+			select {
+			case <-ctx.Done():
+				return
+			case <-sigCh:
+				size, err := term.GetWinsize(inFd)
+				if err != nil {
+					slog.Debug("get window size", "error", err)
+					continue
+				}
+				if err = sendResizeRequest(stream, size); err != nil {
+					slog.Debug("send resize request", "error", err)
+				}
+			}
+		}
+	}()
+
+	return nil
+}
+
+// handleClientInputStream reads from stdin and sends data to the remote server.
+// It also periodically checks for context cancellation to exit gracefully when e.g.
+// the output stream is closed.
+func handleClientInputStream(ctx context.Context, stream pb.Docker_ExecContainerClient, stdin io.Reader) error {
+	slog.Debug("Input goroutine started")
+	defer slog.Debug("Input goroutine exited")
+
+	defer stream.CloseSend()
+
+	// Channel to receive stdin data
+	stdinCh := make(chan []byte)
+
+	stdinErrCh := make(chan error, 1)
+
+	// Read from stdin in a separate goroutine
+	// Note: this goroutine may continue blocking on Read even after we exit from the function,
+	// but that's OK - it will eventually unblock when data arrives or stdin closes.
+	go func() {
+		buf := make([]byte, 32*1024) // 32KB buffer
+		for {
+			n, err := stdin.Read(buf)
+			if n > 0 {
+				data := make([]byte, n)
+				copy(data, buf[:n])
+				select {
+				case stdinCh <- data:
+				case <-ctx.Done():
+					slog.Debug("stdin reader exiting due to context done")
+					return
+				}
+			}
+			if err != nil {
+				if err == io.EOF {
+					slog.Debug("stdin reader: EOF received")
+				} else {
+					slog.Debug("stdin reader error", "error", err)
+				}
+				stdinErrCh <- err
+				return
+			}
+		}
+	}()
+
+	// Send stdin data to the server or exit when context is cancelled
+	for {
+		select {
+		case <-ctx.Done():
+			return nil
+		case data := <-stdinCh:
+			if err := stream.Send(&pb.ExecContainerRequest{
+				Payload: &pb.ExecContainerRequest_Stdin{Stdin: data},
+			}); err != nil {
+				return fmt.Errorf("send stdin: %w", err)
+			}
+		case err := <-stdinErrCh:
+			if err != io.EOF {
+				return fmt.Errorf("read stdin: %w", err)
+			}
+			return nil
+		}
+	}
+}
+
+// handleClientOutputStream receives output from the exec stream and writes to stdout/stderr.
+// It also captures the exit code and signals completion via context cancellation.
+func handleClientOutputStream(ctx context.Context, stream pb.Docker_ExecContainerClient, stdout, stderr io.Writer, exitCode *int) error {
+	slog.Debug("Output goroutine started")
+	defer slog.Debug("Output goroutine exited")
+
+	for {
+		resp, err := stream.Recv()
+		if err == io.EOF {
+			slog.Debug("output stream: EOF received")
+			return nil
+		}
+		if err != nil {
+			return fmt.Errorf("receive from stream: %w", err)
+		}
+
+		switch payload := resp.Payload.(type) {
+		case *pb.ExecContainerResponse_ExecId:
+			// This is sent first; we already processed it earlier, so just ignore duplicates.
+		case *pb.ExecContainerResponse_Stdout:
+			if _, err := stdout.Write(payload.Stdout); err != nil {
+				return fmt.Errorf("write stdout: %w", err)
+			}
+		case *pb.ExecContainerResponse_Stderr:
+			if _, err := stderr.Write(payload.Stderr); err != nil {
+				return fmt.Errorf("write stderr: %w", err)
+			}
+		case *pb.ExecContainerResponse_ExitCode:
+			slog.Debug("received exit code", "code", payload.ExitCode)
+			*exitCode = int(payload.ExitCode)
+			return nil
+		}
+	}
+}
+
+// ExecContainer executes a command in a running container with bidirectional streaming.
+// TODO: This can be merged with pkg/client as it's an unnecessary logic split.
+func (c *Client) ExecContainer(ctx context.Context, opts ExecConfig) (exitCode int, err error) {
+	// Note: In non-interactive mode (without TTY), signals like SIGTERM/SIGINT will terminate the client
+	// process without being forwarded to the remote container process. We could catch and forward these
+	// signals (useful for long-running commands), but for example "kubectl exec" doesn't do this either (as of November 2025),
+	// so we keep it simple for now.
+
+	slog.Debug("starting ExecContainer", "containerID", opts.ContainerID, "options", opts.Options)
+
+	// Initialize exit code to non-zero in case we have to return early
+	exitCode = 1
+
+	// Set up I/O streams - use custom streams if provided, otherwise default to os.Stdin/Stdout/Stderr
+	stdin := io.Reader(os.Stdin)
+	stdout := io.Writer(os.Stdout)
+	stderr := io.Writer(os.Stderr)
+
+	if opts.Options.Stdin != nil {
+		stdin = opts.Options.Stdin
+	}
+	if opts.Options.Stdout != nil {
+		stdout = opts.Options.Stdout
+	}
+	if opts.Options.Stderr != nil {
+		stderr = opts.Options.Stderr
+	}
+
+	// Marshal the exec config
+	configBytes, err := json.Marshal(opts.Options)
+	if err != nil {
+		return -1, fmt.Errorf("marshal exec config: %w", err)
+	}
+
+	// Create the bidirectional stream
+	stream, err := c.GRPCClient.ExecContainer(ctx)
+	if err != nil {
+		return -1, fmt.Errorf("create exec stream: %w", err)
+	}
+
+	// Send the initial configuration
+	if err := stream.Send(&pb.ExecContainerRequest{
+		Payload: &pb.ExecContainerRequest_Config{
+			Config: &pb.ExecConfig{
+				ContainerId: opts.ContainerID,
+				Options:     configBytes,
+			},
+		},
+	}); err != nil {
+		return -1, fmt.Errorf("send exec config: %w", err)
+	}
+
+	// Receive the exec ID
+	resp, err := stream.Recv()
+	if err != nil {
+		return -1, fmt.Errorf("receive exec ID: %w", err)
+	}
+	execID := resp.GetExecId()
+	if execID == "" {
+		return -1, fmt.Errorf("expected exec ID in first response")
+	}
+
+	errGroup, ctx := errgroup.WithContext(ctx)
+
+	// Create cancellable context for goroutine coordination
+	ctx, cancel := context.WithCancel(ctx)
+	defer cancel()
+
+	// Handle terminal setup for interactive sessions
+	if opts.Options.AttachStdin && opts.Options.Tty {
+		restoreTerminal, err := setupTerminal(ctx, stream)
+		if err != nil {
+			return -1, fmt.Errorf("setup terminal: %w", err)
+		}
+		if restoreTerminal != nil {
+			defer restoreTerminal()
+		}
+	}
+
+	// Handle stdin stream if needed
+	if opts.Options.AttachStdin {
+		errGroup.Go(func() error {
+			return handleClientInputStream(ctx, stream, stdin)
+		})
+	} else {
+		// Close send direction immediately if not attaching stdin
+		stream.CloseSend()
+	}
+
+	// Handle output streams (stdout/stderr)
+	errGroup.Go(func() error {
+		defer cancel()
+		return handleClientOutputStream(ctx, stream, stdout, stderr, &exitCode)
+	})
+
+	err = errGroup.Wait()
+
+	if err == nil && opts.Options.Detach {
+		return 0, nil
+	}
+
+	return exitCode, err
+}

internal/machine/docker/controller_linux.go

@@ -8,10 +8,10 @@ import (
 	"net/netip"
 	"strconv"
 
+	"github.com/containerd/errdefs"
 	dockercontainer "github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/filters"
 	dnetwork "github.com/docker/docker/api/types/network"
-	"github.com/docker/docker/client"
 	"github.com/docker/docker/libnetwork/iptables"
 	"github.com/psviderski/uncloud/internal/machine/dns"
 	"github.com/psviderski/uncloud/internal/machine/firewall"
@@ -27,7 +27,7 @@ func (c *Controller) EnsureUncloudNetwork(ctx context.Context, subnet netip.Pref
 	needsCreation := false
 	nw, err := c.client.NetworkInspect(ctx, NetworkName, dnetwork.InspectOptions{})
 	if err != nil {
-		if !client.IsErrNotFound(err) {
+		if !errdefs.IsNotFound(err) {
 			return fmt.Errorf("inspect Docker network '%s': %w", NetworkName, err)
 		}
 		needsCreation = true
@@ -187,7 +187,7 @@ func (c *Controller) Cleanup() error {
 
 		for _, ctr := range containers {
 			err = c.client.ContainerStop(ctx, ctr.ID, dockercontainer.StopOptions{})
-			if err != nil && !client.IsErrNotFound(err) {
+			if err != nil && !errdefs.IsNotFound(err) {
 				errs = append(errs, fmt.Errorf("stop container '%s': %w", ctr.ID, err))
 			}
 
@@ -197,7 +197,7 @@ func (c *Controller) Cleanup() error {
 			})
 			if err == nil {
 				removed++
-			} else if !client.IsErrNotFound(err) {
+			} else if !errdefs.IsNotFound(err) {
 				errs = append(errs, fmt.Errorf("remove container '%s': %w", ctr.ID, err))
 			}
 		}
@@ -223,10 +223,10 @@ func (c *Controller) Cleanup() error {
 
 		if err = c.client.NetworkRemove(ctx, NetworkName); err == nil {
 			slog.Info("Docker network removed.", "name", NetworkName)
-		} else if !client.IsErrNotFound(err) {
+		} else if !errdefs.IsNotFound(err) {
 			errs = append(errs, fmt.Errorf("remove Docker network '%s': %w", NetworkName, err))
 		}
-	} else if !client.IsErrNotFound(err) {
+	} else if !errdefs.IsNotFound(err) {
 		errs = append(errs, fmt.Errorf("inspect Docker network '%s': %w", NetworkName, err))
 	}
 

internal/machine/docker/server.go

@@ -1,19 +1,25 @@
 package docker
 
 import (
+	"archive/tar"
+	"bytes"
 	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
 	"io"
 	"log/slog"
+	"maps"
 	"net/netip"
 	"os"
+	"path/filepath"
 	"regexp"
 	"slices"
 	"strconv"
 	"strings"
+	"time"
 
+	"github.com/containerd/errdefs"
 	"github.com/distribution/reference"
 	dockercommand "github.com/docker/cli/cli/command"
 	dockerconfig "github.com/docker/cli/cli/config"
@@ -25,6 +31,7 @@ import (
 	"github.com/docker/docker/api/types/network"
 	"github.com/docker/docker/api/types/volume"
 	"github.com/docker/docker/client"
+	"github.com/docker/docker/pkg/stdcopy"
 	"github.com/docker/go-connections/nat"
 	"github.com/google/go-containerregistry/pkg/authn"
 	"github.com/google/go-containerregistry/pkg/name"
@@ -40,6 +47,7 @@ import (
 	"google.golang.org/grpc/codes"
 	"google.golang.org/grpc/status"
 	"google.golang.org/protobuf/types/known/emptypb"
+	"google.golang.org/protobuf/types/known/timestamppb"
 )
 
 var fullDockerIDRegex = regexp.MustCompile(`^[a-f0-9]{64}$`)
@@ -53,41 +61,35 @@ type Server struct {
 	// internalDNSIP is a function that returns the IP address of the internal DNS server. It may return an empty
 	// address if the address is unknown (e.g. when the machine is not initialised yet).
 	internalDNSIP func() netip.Addr
+	// machineID is a function that returns the machine ID. It may return an empty string if the machine
+	// is not initialised yet.
+	machineID func() string
 	// networkReady is a function that returns true if the Docker network is ready for containers.
 	networkReady func() bool
 	// waitForNetworkReady is a function that waits for the Docker network to be ready for containers.
 	waitForNetworkReady func(ctx context.Context) error
 }
 
-// ServerOption configures the Docker server.
-type ServerOption func(*Server)
-
-// WithNetworkReady sets the network readiness check function.
-func WithNetworkReady(networkReady func() bool) ServerOption {
-	return func(s *Server) {
-		s.networkReady = networkReady
-	}
-}
-
-// WithWaitForNetworkReady sets the network readiness wait function.
-func WithWaitForNetworkReady(waitForNetworkReady func(ctx context.Context) error) ServerOption {
-	return func(s *Server) {
-		s.waitForNetworkReady = waitForNetworkReady
-	}
+type ServerOptions struct {
+	// TODO: verify if we still need the network readiness checks as the cluster controller ensures the network
+	//  is ready before starting the network API server. It may still be needed when communicating with the local
+	//  API server but in this case we should probably fail until the cluster is initialised.
+	NetworkReady        func() bool
+	WaitForNetworkReady func(ctx context.Context) error
 }
 
 // NewServer creates a new Docker gRPC server with the provided Docker service.
-func NewServer(service *Service, db *sqlx.DB, internalDNSIP func() netip.Addr, opts ...ServerOption) *Server {
+func NewServer(service *Service, db *sqlx.DB, internalDNSIP func() netip.Addr, machineID func() string, opts ServerOptions) *Server {
 	s := &Server{
 		client:        service.Client,
 		service:       service,
 		db:            db,
 		internalDNSIP: internalDNSIP,
+		machineID:     machineID,
 	}
 
-	for _, opt := range opts {
-		opt(s)
-	}
+	s.networkReady = opts.NetworkReady
+	s.waitForNetworkReady = opts.WaitForNetworkReady
 
 	return s
 }
@@ -115,10 +117,10 @@ func (s *Server) CreateContainer(ctx context.Context, req *pb.CreateContainerReq
 
 	resp, err := s.client.ContainerCreate(ctx, &config, &hostConfig, &networkConfig, &platform, req.Name)
 	if err != nil {
-		if client.IsErrNotFound(err) {
-			return nil, status.Errorf(codes.NotFound, err.Error())
+		if errdefs.IsNotFound(err) {
+			return nil, status.Error(codes.NotFound, err.Error())
 		}
-		return nil, status.Errorf(codes.Internal, err.Error())
+		return nil, status.Error(codes.Internal, err.Error())
 	}
 
 	respBytes, err := json.Marshal(resp)
@@ -133,10 +135,10 @@ func (s *Server) CreateContainer(ctx context.Context, req *pb.CreateContainerReq
 func (s *Server) InspectContainer(ctx context.Context, req *pb.InspectContainerRequest) (*pb.InspectContainerResponse, error) {
 	resp, err := s.client.ContainerInspect(ctx, req.Id)
 	if err != nil {
-		if client.IsErrNotFound(err) {
-			return nil, status.Errorf(codes.NotFound, err.Error())
+		if errdefs.IsNotFound(err) {
+			return nil, status.Error(codes.NotFound, err.Error())
 		}
-		return nil, status.Errorf(codes.Internal, err.Error())
+		return nil, status.Error(codes.Internal, err.Error())
 	}
 
 	respBytes, err := json.Marshal(resp)
@@ -166,10 +168,10 @@ func (s *Server) StartContainer(ctx context.Context, req *pb.StartContainerReque
 	}
 
 	if err := s.client.ContainerStart(ctx, req.Id, opts); err != nil {
-		if client.IsErrNotFound(err) {
-			return nil, status.Errorf(codes.NotFound, err.Error())
+		if errdefs.IsNotFound(err) {
+			return nil, status.Error(codes.NotFound, err.Error())
 		}
-		return nil, status.Errorf(codes.Internal, err.Error())
+		return nil, status.Error(codes.Internal, err.Error())
 	}
 
 	return &emptypb.Empty{}, nil
@@ -185,10 +187,10 @@ func (s *Server) StopContainer(ctx context.Context, req *pb.StopContainerRequest
 	}
 
 	if err := s.client.ContainerStop(ctx, req.Id, opts); err != nil {
-		if client.IsErrNotFound(err) {
-			return nil, status.Errorf(codes.NotFound, err.Error())
+		if errdefs.IsNotFound(err) {
+			return nil, status.Error(codes.NotFound, err.Error())
 		}
-		return nil, status.Errorf(codes.Internal, err.Error())
+		return nil, status.Error(codes.Internal, err.Error())
 	}
 
 	return &emptypb.Empty{}, nil
@@ -218,13 +220,13 @@ func (s *Server) ListContainers(ctx context.Context, req *pb.ListContainersReque
 
 	containerSummaries, err := s.client.ContainerList(ctx, opts)
 	if err != nil {
-		return nil, status.Errorf(codes.Internal, err.Error())
+		return nil, status.Error(codes.Internal, err.Error())
 	}
-	containers := make([]types.ContainerJSON, 0, len(containerSummaries))
+	containers := make([]container.InspectResponse, 0, len(containerSummaries))
 	for _, cs := range containerSummaries {
 		c, err := s.client.ContainerInspect(ctx, cs.ID)
 		if err != nil {
-			if client.IsErrNotFound(err) {
+			if errdefs.IsNotFound(err) {
 				// The listed container may have been removed while we were inspecting other containers.
 				continue
 			}
@@ -257,10 +259,10 @@ func (s *Server) RemoveContainer(ctx context.Context, req *pb.RemoveContainerReq
 	}
 
 	if err := s.client.ContainerRemove(ctx, req.Id, opts); err != nil {
-		if client.IsErrNotFound(err) {
-			return nil, status.Errorf(codes.NotFound, err.Error())
+		if errdefs.IsNotFound(err) {
+			return nil, status.Error(codes.NotFound, err.Error())
 		}
-		return nil, status.Errorf(codes.Internal, err.Error())
+		return nil, status.Error(codes.Internal, err.Error())
 	}
 
 	return &emptypb.Empty{}, nil
@@ -286,7 +288,7 @@ func (s *Server) PullImage(req *pb.PullImageRequest, stream grpc.ServerStreaming
 
 	respBody, err := s.client.ImagePull(ctx, req.Image, opts)
 	if err != nil {
-		return status.Errorf(codes.Internal, err.Error())
+		return status.Error(codes.Internal, err.Error())
 	}
 	defer respBody.Close()
 
@@ -317,19 +319,19 @@ func (s *Server) PullImage(req *pb.PullImageRequest, stream grpc.ServerStreaming
 		case err = <-errCh:
 			return err
 		case <-ctx.Done():
-			return status.Errorf(codes.Canceled, ctx.Err().Error())
+			return status.Error(codes.Canceled, ctx.Err().Error())
 		}
 	}
 }
 
 // InspectImage returns the image information for the given image ID.
 func (s *Server) InspectImage(ctx context.Context, req *pb.InspectImageRequest) (*pb.InspectImageResponse, error) {
-	resp, _, err := s.client.ImageInspectWithRaw(ctx, req.Id)
+	resp, err := s.client.ImageInspect(ctx, req.Id)
 	if err != nil {
-		if client.IsErrNotFound(err) {
-			return nil, status.Errorf(codes.NotFound, err.Error())
+		if errdefs.IsNotFound(err) {
+			return nil, status.Error(codes.NotFound, err.Error())
 		}
-		return nil, status.Errorf(codes.Internal, err.Error())
+		return nil, status.Error(codes.Internal, err.Error())
 	}
 
 	respBytes, err := json.Marshal(resp)
@@ -385,6 +387,49 @@ func (s *Server) InspectRemoteImage(
 	}, nil
 }
 
+// ListImages returns a list of all images matching the filter and indicates whether Docker is using the containerd
+// image store.
+func (s *Server) ListImages(ctx context.Context, req *pb.ListImagesRequest) (*pb.ListImagesResponse, error) {
+	var opts image.ListOptions
+	if len(req.Options) > 0 {
+		if err := json.Unmarshal(req.Options, &opts); err != nil {
+			return nil, status.Errorf(codes.InvalidArgument, "unmarshal options: %v", err)
+		}
+
+		// Handle filters separately because they implement custom JSON unmarshalling.
+		var raw map[string]json.RawMessage
+		if err := json.Unmarshal(req.Options, &raw); err != nil {
+			return nil, status.Errorf(codes.InvalidArgument, "unmarshal options to raw map: %v", err)
+		}
+
+		if filtersBytes, ok := raw["Filters"]; ok {
+			args, err := filters.FromJSON(string(filtersBytes))
+			if err != nil {
+				return nil, status.Errorf(codes.InvalidArgument, "unmarshal filters: %v", err)
+			}
+			opts.Filters = args
+		}
+	}
+
+	images, err := s.service.ListImages(ctx, opts)
+	if err != nil {
+		return nil, status.Error(codes.Internal, err.Error())
+	}
+
+	machineImages := pb.MachineImages{
+		ContainerdStore: images.ContainerdStore,
+	}
+	if len(images.Images) > 0 {
+		if machineImages.Images, err = json.Marshal(images.Images); err != nil {
+			return nil, status.Errorf(codes.Internal, "marshal Docker images: %v", err)
+		}
+	}
+
+	return &pb.ListImagesResponse{
+		Messages: []*pb.MachineImages{&machineImages},
+	}, nil
+}
+
 // CreateVolume creates a new volume with the given options.
 func (s *Server) CreateVolume(ctx context.Context, req *pb.CreateVolumeRequest) (*pb.CreateVolumeResponse, error) {
 	var opts volume.CreateOptions
@@ -456,7 +501,7 @@ func (s *Server) ListVolumes(ctx context.Context, req *pb.ListVolumesRequest) (*
 // RemoveVolume removes a volume with the given ID.
 func (s *Server) RemoveVolume(ctx context.Context, req *pb.RemoveVolumeRequest) (*emptypb.Empty, error) {
 	if err := s.client.VolumeRemove(ctx, req.Id, req.Force); err != nil {
-		if client.IsErrNotFound(err) {
+		if errdefs.IsNotFound(err) {
 			return nil, status.Error(codes.NotFound, err.Error())
 		}
 		return nil, status.Error(codes.Internal, err.Error())
@@ -492,9 +537,21 @@ func (s *Server) CreateServiceContainer(
 		containerName = fmt.Sprintf("%s-%s", spec.Name, suffix)
 	}
 
+	envVars := maps.Clone(spec.Container.Env)
+	if envVars == nil {
+		envVars = make(api.EnvVars)
+	}
+
+	// Inject the machine ID if available
+	if s.machineID != nil {
+		if machineID := s.machineID(); machineID != "" {
+			envVars["UNCLOUD_MACHINE_ID"] = machineID
+		}
+	}
+
 	config := &container.Config{
 		Cmd:        spec.Container.Command,
-		Env:        spec.Container.Env.ToSlice(),
+		Env:        envVars.ToSlice(),
 		Entrypoint: spec.Container.Entrypoint,
 		Hostname:   containerName,
 		Image:      spec.Container.Image,
@@ -557,6 +614,7 @@ func (s *Server) CreateServiceContainer(
 			NanoCPUs:          spec.Container.Resources.CPU,
 			Memory:            spec.Container.Resources.Memory,
 			MemoryReservation: spec.Container.Resources.MemoryReservation,
+			DeviceRequests:    spec.Container.Resources.DeviceReservations,
 		},
 		// Restart service containers if they exit or a machine restarts unless they are explicitly stopped.
 		// For one-off containers and batch jobs we plan to use a different service type/mode.
@@ -590,12 +648,19 @@ func (s *Server) CreateServiceContainer(
 
 	resp, err := s.client.ContainerCreate(ctx, config, hostConfig, networkConfig, nil, containerName)
 	if err != nil {
-		if client.IsErrNotFound(err) {
+		if errdefs.IsNotFound(err) {
 			return nil, status.Error(codes.NotFound, err.Error())
 		}
 		return nil, status.Error(codes.Internal, err.Error())
 	}
 
+	// Inject configs into the created container
+	if err = s.injectConfigs(ctx, resp.ID, spec.Configs, spec.Container.ConfigMounts); err != nil {
+		// Remove the container if config injection fails
+		_ = s.client.ContainerRemove(ctx, resp.ID, container.RemoveOptions{RemoveVolumes: true})
+		return nil, status.Errorf(codes.Internal, "inject configs: %v", err)
+	}
+
 	respBytes, err := json.Marshal(resp)
 	if err != nil {
 		return nil, status.Errorf(codes.Internal, "marshal response: %v", err)
@@ -672,6 +737,122 @@ func ToDockerMounts(volumes []api.VolumeSpec, mounts []api.VolumeMount) ([]mount
 	return dockerMounts, nil
 }
 
+// injectConfigs writes config content directly into the container.
+// It processes ConfigSpecs and ConfigMounts to mount configuration content into the container filesystem.
+func (s *Server) injectConfigs(ctx context.Context, containerID string, configs []api.ConfigSpec, mounts []api.ConfigMount) error {
+	if len(configs) == 0 || len(mounts) == 0 {
+		return nil
+	}
+
+	if err := api.ValidateConfigsAndMounts(configs, mounts); err != nil {
+		return fmt.Errorf("validate configs and mounts: %w", err)
+	}
+
+	// Create a map of config name to config spec for quick lookup
+	configMap := make(map[string]api.ConfigSpec)
+	for _, config := range configs {
+		configMap[config.Name] = config
+	}
+
+	// Process each config mount
+	for _, m := range mounts {
+		config, exists := configMap[m.ConfigName]
+		if !exists {
+			return fmt.Errorf("config mount references a config that doesn't exist: '%s'", m.ConfigName)
+		}
+
+		// Determine target path in container
+		targetPath := m.ContainerPath
+		if targetPath == "" {
+			// This is the default from the Compose spec
+			targetPath = filepath.Join("/", m.ConfigName)
+		}
+
+		// Determine file mode
+		fileMode := os.FileMode(0o444) // Default permissions
+		if m.Mode != nil {
+			fileMode = *m.Mode
+		}
+
+		uid, err := m.GetNumericUid()
+		if err != nil {
+			return fmt.Errorf("invalid Uid: %w", err)
+		}
+
+		gid, err := m.GetNumericGid()
+		if err != nil {
+			return fmt.Errorf("invalid Gid: %w", err)
+		}
+
+		// Copy the config content directly into the container
+		if err := s.copyContentToContainer(
+			ctx, containerID, config.Content, targetPath, uid, gid, fileMode,
+		); err != nil {
+			return fmt.Errorf("copy config file '%s' to container: %w", config.Name, err)
+		}
+
+		slog.Debug("Injected config into container",
+			"config", config.Name,
+			"container", containerID[:12],
+			"target", targetPath)
+	}
+
+	return nil
+}
+
+// copyContentToContainer copies content directly to a file in the container using Docker's CopyToContainer API.
+func (s *Server) copyContentToContainer(ctx context.Context, containerID string, content []byte, targetPath string, uid *uint64, gid *uint64, fileMode os.FileMode) error {
+	// Create a tar archive containing the file
+	var buf bytes.Buffer
+	tw := tar.NewWriter(&buf)
+
+	// Create tar header
+	header := &tar.Header{
+		Name:     filepath.Base(targetPath),
+		Size:     int64(len(content)),
+		Mode:     int64(fileMode),
+		ModTime:  time.Now(),
+		Typeflag: tar.TypeReg,
+	}
+
+	// Set ownership if specified
+	if uid != nil {
+		header.Uid = int(*uid)
+	}
+	if gid != nil {
+		header.Gid = int(*gid)
+	}
+
+	// Write header and content to tar archive
+	if err := tw.WriteHeader(header); err != nil {
+		return fmt.Errorf("write tar header: %w", err)
+	}
+	if _, err := tw.Write(content); err != nil {
+		return fmt.Errorf("write content to tar: %w", err)
+	}
+	if err := tw.Close(); err != nil {
+		return fmt.Errorf("close tar writer: %w", err)
+	}
+
+	// Copy the tar archive to the container
+	targetDir := filepath.Dir(targetPath)
+	if targetDir == "." {
+		targetDir = "/"
+	}
+
+	if err := s.client.CopyToContainer(
+		ctx,
+		containerID,
+		targetDir,
+		&buf,
+		container.CopyToContainerOptions{CopyUIDGID: true},
+	); err != nil {
+		return fmt.Errorf("copy to container: %w", err)
+	}
+
+	return nil
+}
+
 func toDockerBindOptions(opts *api.BindOptions) *mount.BindOptions {
 	if opts == nil {
 		return nil
@@ -703,7 +884,7 @@ func (s *Server) verifyDockerVolumesExist(ctx context.Context, mounts []mount.Mo
 
 		// TODO: non-local volume drivers should likely be handled differently (needs proper investigation).
 		if _, err := s.client.VolumeInspect(ctx, m.Source); err != nil {
-			if client.IsErrNotFound(err) {
+			if errdefs.IsNotFound(err) {
 				return status.Errorf(codes.NotFound, "volume '%s' not found", m.Source)
 			}
 			return status.Errorf(codes.Internal, "inspect volume '%s': %v", m.Source, err.Error())
@@ -723,10 +904,10 @@ func (s *Server) InspectServiceContainer(
 ) (*pb.ServiceContainer, error) {
 	serviceCtr, err := s.service.InspectServiceContainer(ctx, req.Id)
 	if err != nil {
-		if client.IsErrNotFound(err) {
-			return nil, status.Errorf(codes.NotFound, err.Error())
+		if errdefs.IsNotFound(err) {
+			return nil, status.Error(codes.NotFound, err.Error())
 		}
-		return nil, status.Errorf(codes.Internal, err.Error())
+		return nil, status.Error(codes.Internal, err.Error())
 	}
 
 	ctrBytes, err := json.Marshal(serviceCtr.Container)
@@ -813,7 +994,7 @@ func (s *Server) RemoveServiceContainer(ctx context.Context, req *pb.RemoveConta
 	if !fullDockerIDRegex.MatchString(req.Id) {
 		ctr, err := s.client.ContainerInspect(ctx, req.Id)
 		if err != nil {
-			if client.IsErrNotFound(err) {
+			if errdefs.IsNotFound(err) {
 				return nil, status.Error(codes.NotFound, err.Error())
 			}
 			return nil, status.Error(codes.Internal, err.Error())
@@ -834,3 +1015,349 @@ func (s *Server) RemoveServiceContainer(ctx context.Context, req *pb.RemoveConta
 
 	return resp, nil
 }
+
+// logsHeartbeatInterval is the interval at which heartbeat entries are sent when there are no logs to stream.
+const logsHeartbeatInterval = 200 * time.Millisecond
+
+// ContainerLogs streams logs from a container.
+func (s *Server) ContainerLogs(
+	req *pb.ContainerLogsRequest, stream grpc.ServerStreamingServer[pb.ContainerLogEntry],
+) error {
+	// Stream context is cancelled when the client has disconnected or the stream has ended.
+	ctx := stream.Context()
+
+	opts := ContainerLogsOptions{
+		ContainerID: req.ContainerId,
+		Follow:      req.Follow,
+		Tail:        int(req.Tail),
+		Since:       req.Since,
+		Until:       req.Until,
+	}
+
+	logsCh, err := s.service.ContainerLogs(ctx, opts)
+	if err != nil {
+		if errdefs.IsNotFound(err) {
+			return status.Error(codes.NotFound, err.Error())
+		}
+		return status.Errorf(codes.Internal, "get container logs: %v", err)
+	}
+
+	log := slog.With("container_id", req.ContainerId, "stream_id", fmt.Sprintf("%p", stream)[2:])
+	log.Debug("Starting container logs streaming.",
+		"follow", req.Follow, "tail", req.Tail, "since", req.Since, "until", req.Until)
+
+	// Heartbeats are needed only when following logs to let the client know when there are no new log entries
+	// to allow it to advance the watermark of last received log timestamp.
+	var heartbeatCh <-chan time.Time
+	if req.Follow {
+		heartbeatTicker := time.NewTicker(logsHeartbeatInterval)
+		defer heartbeatTicker.Stop()
+		heartbeatCh = heartbeatTicker.C
+	}
+
+	started := time.Now()
+	lastSent := time.Time{}
+
+	for {
+		select {
+		case entry, ok := <-logsCh:
+			if !ok {
+				// Channel closed, no more log entries.
+				return nil
+			}
+
+			if entry.Err != nil {
+				return status.Error(codes.Internal, entry.Err.Error())
+			}
+
+			pbEntry := &pb.ContainerLogEntry{
+				Stream:    api.LogStreamTypeToProto(entry.Stream),
+				Timestamp: timestamppb.New(entry.Timestamp),
+				Message:   entry.Message,
+			}
+			if err = stream.Send(pbEntry); err != nil {
+				return status.Errorf(codes.Internal, "send log entry: %v", err)
+			}
+			lastSent = entry.Timestamp
+
+		case now := <-heartbeatCh:
+			// Only send heartbeat if no log entries have been sent since the last heartbeat interval or
+			// if no log entries have been sent at all for at least a heartbeat interval since starting.
+			if now.Sub(lastSent) < logsHeartbeatInterval ||
+				(lastSent.IsZero() && now.Sub(started) < logsHeartbeatInterval) {
+				continue
+			}
+
+			// Use the timestamp one heartbeat in the past to be conservative. This reduces the chance of sending
+			// a timestamp that is greater than a log entry currently being parsed but not yet sent, which would
+			// cause the client to incorrectly believe it has received all logs up to that point.
+			heartbeat := &pb.ContainerLogEntry{
+				Stream:    pb.ContainerLogEntry_HEARTBEAT,
+				Timestamp: timestamppb.New(now.Add(-logsHeartbeatInterval)),
+			}
+			if err = stream.Send(heartbeat); err != nil {
+				return status.Errorf(codes.Internal, "send log stream heartbeat: %v", err)
+			}
+			lastSent = heartbeat.Timestamp.AsTime()
+			log.Debug("Sent log stream heartbeat.", "timestamp", lastSent)
+
+		case <-ctx.Done():
+			return status.Error(codes.Canceled, ctx.Err().Error())
+		}
+	}
+}
+
+// receiveExecConfig receives and validates the initial exec configuration from the stream.
+func (s *Server) receiveExecConfig(stream pb.Docker_ExecContainerServer) (*pb.ExecConfig, api.ExecOptions, error) {
+	req, err := stream.Recv()
+	if err != nil {
+		return nil, api.ExecOptions{}, status.Errorf(codes.InvalidArgument, "receive config: %v", err)
+	}
+
+	execConfig := req.GetConfig()
+	if execConfig == nil {
+		return nil, api.ExecOptions{}, status.Error(codes.InvalidArgument, "first message must contain exec config")
+	}
+
+	// Unmarshal the Uncloud's execOpts
+	var execOpts api.ExecOptions
+	if err := json.Unmarshal(execConfig.Options, &execOpts); err != nil {
+		return nil, api.ExecOptions{}, status.Errorf(codes.InvalidArgument, "unmarshal exec config: %v", err)
+	}
+
+	return execConfig, execOpts, nil
+}
+
+// handleServerExecInput reads from the gRPC stream and writes to Docker stdin, handling resize requests.
+func (s *Server) handleServerExecInput(
+	ctx context.Context,
+	stream pb.Docker_ExecContainerServer,
+	attachConn types.HijackedResponse,
+	execID string,
+	tty bool,
+) error {
+	slog.Debug("Input goroutine started", "exec_id", execID, "tty", tty)
+	defer slog.Debug("Input goroutine exited", "exec_id", execID)
+
+	defer attachConn.CloseWrite()
+	for {
+		select {
+		case <-ctx.Done():
+			slog.Debug("Input goroutine context canceled via context", "exec_id", execID)
+			return nil
+		default:
+		}
+
+		req, err := stream.Recv()
+		switch {
+		case errors.Is(err, io.EOF):
+			slog.Debug("Input goroutine received EOF", "exec_id", execID)
+			return nil
+		case status.Code(err) == codes.Canceled:
+			// Can be the case when the output goroutine ends and the stream context is canceled.
+			slog.Debug("Input goroutine context canceled", "exec_id", execID)
+			return nil
+		case err == nil:
+			// continue processing
+		default:
+			return fmt.Errorf("receive from stream: %w", err)
+		}
+
+		switch payload := req.Payload.(type) {
+		case *pb.ExecContainerRequest_Stdin:
+			if _, err := attachConn.Conn.Write(payload.Stdin); err != nil {
+				return fmt.Errorf("write to stdin: %w", err)
+			}
+		case *pb.ExecContainerRequest_Resize:
+			if tty {
+				resizeOpts := container.ResizeOptions{
+					Height: uint(payload.Resize.Height),
+					Width:  uint(payload.Resize.Width),
+				}
+				if err := s.client.ContainerExecResize(ctx, execID, resizeOpts); err != nil {
+					slog.Warn("Failed to resize TTY", "err", err, "exec_id", execID)
+				}
+			}
+		}
+	}
+}
+
+// handleServerExecOutput reads from Docker stdout/stderr and writes to the gRPC stream.
+func (s *Server) handleServerExecOutput(
+	stream pb.Docker_ExecContainerServer,
+	attachResp types.HijackedResponse,
+	execID string,
+	tty bool,
+) error {
+	slog.Debug("Output goroutine started", "exec_id", execID, "tty", tty)
+	defer slog.Debug("Output goroutine exited", "exec_id", execID)
+
+	if tty {
+		// In TTY mode, all output is stdout - copy directly to stream
+		stdoutWriter := &grpcStreamWriter{stream: stream, isStderr: false}
+		_, err := io.Copy(stdoutWriter, attachResp.Reader)
+		if err != nil && err != io.EOF {
+			return fmt.Errorf("copy tty output: %w", err)
+		}
+		return nil
+	} else {
+		// In non-TTY mode, Docker multiplexes stdout/stderr with headers
+		// Use stdcopy to demultiplex
+		slog.Debug("Starting StdCopy for non-TTY", "exec_id", execID)
+		stdoutWriter := &grpcStreamWriter{stream: stream, isStderr: false}
+		stderrWriter := &grpcStreamWriter{stream: stream, isStderr: true}
+
+		written, err := stdcopy.StdCopy(stdoutWriter, stderrWriter, attachResp.Reader)
+		slog.Debug("StdCopy completed", "exec_id", execID, "bytes", written, "err", err)
+		if err != nil && err != io.EOF {
+			return fmt.Errorf("demultiplex docker output: %w", err)
+		}
+		return nil
+	}
+}
+
+// grpcStreamWriter is a writer that sends data to a gRPC stream as stdout or stderr.
+type grpcStreamWriter struct {
+	stream   pb.Docker_ExecContainerServer
+	isStderr bool
+}
+
+func (w *grpcStreamWriter) Write(p []byte) (n int, err error) {
+	data := make([]byte, len(p))
+	copy(data, p)
+
+	var resp *pb.ExecContainerResponse
+	if w.isStderr {
+		resp = &pb.ExecContainerResponse{
+			Payload: &pb.ExecContainerResponse_Stderr{Stderr: data},
+		}
+	} else {
+		resp = &pb.ExecContainerResponse{
+			Payload: &pb.ExecContainerResponse_Stdout{Stdout: data},
+		}
+	}
+
+	if err := w.stream.Send(resp); err != nil {
+		return 0, err
+	}
+	return len(p), nil
+}
+
+// ExecContainer executes a command in a running container with bidirectional streaming for stdin/stdout/stderr.
+func (s *Server) ExecContainer(stream pb.Docker_ExecContainerServer) error {
+	slog.Debug("ExecContainer server-side called")
+	defer slog.Debug("ExecContainer server-side ended")
+
+	ctx := stream.Context()
+
+	// Receive and validate configuration
+	execConfig, execOpts, err := s.receiveExecConfig(stream)
+	if err != nil {
+		return err
+	}
+
+	// Convert to Docker's ExecOptions
+	dockerExecOpts := container.ExecOptions{
+		Cmd:          execOpts.Command,
+		AttachStdin:  execOpts.AttachStdin,
+		AttachStdout: execOpts.AttachStdout,
+		AttachStderr: execOpts.AttachStderr,
+		Tty:          execOpts.Tty,
+		User:         execOpts.User,
+		Privileged:   execOpts.Privileged,
+		WorkingDir:   execOpts.WorkingDir,
+		Env:          execOpts.Env,
+	}
+
+	// Create the exec instance
+	execResp, err := s.client.ContainerExecCreate(ctx, execConfig.ContainerId, dockerExecOpts)
+	if err != nil {
+		if errdefs.IsNotFound(err) {
+			return status.Error(codes.NotFound, err.Error())
+		}
+		return status.Errorf(codes.Internal, "create exec: %v", err)
+	}
+
+	// Send the exec ID back to the client
+	if err := stream.Send(&pb.ExecContainerResponse{
+		Payload: &pb.ExecContainerResponse_ExecId{ExecId: execResp.ID},
+	}); err != nil {
+		return status.Errorf(codes.Internal, "send exec ID: %v", err)
+	}
+	slog.Debug("Sent exec ID to the client", "exec_id", execResp.ID)
+
+	// For detached mode, start without attaching
+	if execOpts.Detach {
+		dockerStartOpts := container.ExecStartOptions{
+			Tty:    dockerExecOpts.Tty,
+			Detach: true,
+		}
+		if err := s.client.ContainerExecStart(ctx, execResp.ID, dockerStartOpts); err != nil {
+			return status.Errorf(codes.Internal, "start exec: %v", err)
+		}
+		return nil
+	}
+
+	// For attached mode, attach to the exec instance
+	attachOpts := container.ExecAttachOptions{
+		Tty: dockerExecOpts.Tty,
+	}
+	attachConn, err := s.client.ContainerExecAttach(ctx, execResp.ID, attachOpts)
+	if err != nil {
+		return status.Errorf(codes.Internal, "attach to exec: %v", err)
+	}
+	defer attachConn.Close()
+
+	// Create a cancelable context for the input handler
+	handlerCtx, cancelInput := context.WithCancel(ctx)
+	defer cancelInput()
+
+	// Create a channel to wait for output completion
+	outputDone := make(chan error, 1)
+
+	// Start stdin handler if stdin is attached
+	if dockerExecOpts.AttachStdin {
+		go func() {
+			err := s.handleServerExecInput(handlerCtx, stream, attachConn, execResp.ID, dockerExecOpts.Tty)
+			if err != nil {
+				slog.Warn("Error in exec input handler", "err", err, "exec_id", execResp.ID)
+			}
+		}()
+	} else {
+		// If not attaching stdin, close the write side immediately
+		attachConn.CloseWrite()
+	}
+
+	// Start output handler
+	// We only wait for this goroutine to complete - it signals when the exec process finishes
+	go func() {
+		outputDone <- s.handleServerExecOutput(stream, attachConn, execResp.ID, dockerExecOpts.Tty)
+	}()
+
+	// Wait for the output goroutine to complete (it signals when done)
+	// We only wait for the output handler goroutine, not for the stdin one.
+	if err := <-outputDone; err != nil {
+		slog.Warn("Error in exec output handler", "err", err, "exec_id", execResp.ID)
+	}
+	// Do a best-effort cancellation of the stdin handler.
+	// We can't guarantee immediate exit because it may be blocked on stream.Recv(), but at
+	// least we want to send a cancel signal explicitly.
+	cancelInput()
+
+	inspectResp, err := s.client.ContainerExecInspect(ctx, execResp.ID)
+	if err != nil {
+		slog.Error("Failed to inspect exec after completion", "err", err, "exec_id", execResp.ID)
+		return status.Errorf(codes.Internal, "inspect exec: %v", err)
+	}
+
+	// Send the exit code
+	slog.Debug("Sending exec exit code", "exec_id", execResp.ID, "exit_code", inspectResp.ExitCode)
+	if err := stream.Send(&pb.ExecContainerResponse{
+		Payload: &pb.ExecContainerResponse_ExitCode{ExitCode: int32(inspectResp.ExitCode)},
+	}); err != nil {
+		slog.Error("Failed to send exec exit code", "err", err, "exec_id", execResp.ID)
+		return status.Errorf(codes.Internal, "send exit code: %v", err)
+	}
+
+	return nil
+}

internal/machine/docker/service.go

@@ -1,24 +1,34 @@
 package docker
 
 import (
+	"bytes"
 	"context"
 	"database/sql"
 	"encoding/json"
 	"errors"
 	"fmt"
 	"log/slog"
+	"strconv"
+	"strings"
+	"time"
 
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/filters"
+	"github.com/docker/docker/api/types/image"
 	"github.com/docker/docker/client"
+	"github.com/docker/docker/pkg/stdcopy"
 	"github.com/jmoiron/sqlx"
 	"github.com/psviderski/uncloud/pkg/api"
+	"google.golang.org/grpc/codes"
+	"google.golang.org/grpc/status"
 )
 
 // Service provides higher-level Docker operations that extends Docker API with Uncloud-specific data
 // from the machine database.
 type Service struct {
+	// Client is a Docker client for managing Docker resources.
 	Client *client.Client
+	// db is a connection to the machine database.
 	db *sqlx.DB
 }
 
@@ -43,7 +53,7 @@ func (s *Service) InspectServiceContainer(ctx context.Context, nameOrID string)
 		return serviceCtr, fmt.Errorf("container '%s' is not managed by Uncloud", nameOrID)
 	}
 
-	serviceCtr.Container = api.Container{ContainerJSON: ctr}
+	serviceCtr.Container = api.Container{InspectResponse: ctr}
 
 	// Retrieve ServiceSpec from the machine database.
 	var specBytes []byte
@@ -101,3 +111,147 @@ func (s *Service) ListServiceContainers(
 
 	return containers, nil
 }
+
+// IsContainerdImageStoreEnabled checks if Docker is configured to use the containerd image store:
+// https://docs.docker.com/engine/storage/containerd/
+func (s *Service) IsContainerdImageStoreEnabled(ctx context.Context) (bool, error) {
+	info, err := s.Client.Info(ctx)
+	if err != nil {
+		return false, fmt.Errorf("get Docker info: %w", err)
+	}
+
+	return strings.Contains(fmt.Sprintf("%s", info.DriverStatus), "containerd.snapshotter"), nil
+}
+
+type Images struct {
+	// Images is a list of images present in the Docker image store (either internal or containerd).
+	Images []image.Summary
+	// ContainerdStore indicates whether Docker is using the containerd image store.
+	ContainerdStore bool
+}
+
+// ListImages lists Docker images with the given options and indicates whether Docker is using the containerd
+// image store. It always includes image manifests in the response if the store is containerd.
+func (s *Service) ListImages(ctx context.Context, opts image.ListOptions) (Images, error) {
+	var imagesResp Images
+
+	// Always include the image manifests in the response.
+	opts.Manifests = true
+	images, err := s.Client.ImageList(ctx, opts)
+	if err != nil {
+		return imagesResp, status.Errorf(codes.Internal, "list images: %v", err)
+	}
+
+	isContainerdStore, err := s.IsContainerdImageStoreEnabled(ctx)
+	if err != nil {
+		return imagesResp, status.Errorf(codes.Internal, "check if Docker uses containerd image store: %v", err)
+	}
+
+	imagesResp = Images{
+		Images:          images,
+		ContainerdStore: isContainerdStore,
+	}
+
+	return imagesResp, nil
+}
+
+// ContainerLogsOptions specifies parameters for ContainerLogs.
+type ContainerLogsOptions struct {
+	ContainerID string
+	Follow      bool
+	Tail        int
+	Since       string
+	Until       string
+}
+
+// ContainerLogs streams logs from a container and returns demultiplexed entries via a channel.
+// The channel is closed when streaming completes or context is cancelled.
+func (s *Service) ContainerLogs(ctx context.Context, opts ContainerLogsOptions) (<-chan api.ContainerLogEntry, error) {
+	dockerOpts := container.LogsOptions{
+		ShowStdout: true,
+		ShowStderr: true,
+		Follow:     opts.Follow,
+		Tail:       strconv.FormatInt(int64(opts.Tail), 10),
+		Since:      opts.Since,
+		Until:      opts.Until,
+		Timestamps: true,
+	}
+
+	reader, err := s.Client.ContainerLogs(ctx, opts.ContainerID, dockerOpts)
+	if err != nil {
+		return nil, err
+	}
+
+	outCh := make(chan api.ContainerLogEntry)
+	stdoutWriter := &logsChannelWriter{ctx: ctx, ch: outCh, isStderr: false}
+	stderrWriter := &logsChannelWriter{ctx: ctx, ch: outCh, isStderr: true}
+
+	// Wrap the context in a cancellable one to unblock the second goroutine below when StdCopy completes.
+	ctx, cancel := context.WithCancel(ctx)
+
+	// Run StdCopy in a goroutine to be able to handle context cancellation.
+	go func() {
+		defer close(outCh)
+		defer cancel()
+
+		// StdCopy is blocking and will return when the reader is closed in another goroutine below or on error.
+		if _, err := stdcopy.StdCopy(stdoutWriter, stderrWriter, reader); err != nil {
+			// Send error as the last entry.
+			select {
+			case outCh <- api.ContainerLogEntry{Err: fmt.Errorf("demultiplex container logs: %w", err)}:
+			case <-ctx.Done():
+			}
+		}
+	}()
+
+	// Close the reader when the context is done to cancel StdCopy if it's still running.
+	go func() {
+		<-ctx.Done()
+		reader.Close()
+	}()
+
+	return outCh, nil
+}
+
+// logsChannelWriter is a writer for stdcopy.StdCopy that sends demultiplexed container logs to a channel.
+type logsChannelWriter struct {
+	ctx      context.Context
+	ch       chan<- api.ContainerLogEntry
+	isStderr bool
+}
+
+func (w *logsChannelWriter) Write(data []byte) (n int, err error) {
+	// Parse timestamp and message from the demultiplexed Docker log payload if the data looks like it contains one.
+	// Format: 2025-01-01T00:00:00.000000000Z message
+	timestamp := time.Time{}
+	message := data
+	if len(data) > 30 && data[4] == '-' && data[7] == '-' && data[10] == 'T' {
+		timestampPart, messagePart, found := bytes.Cut(data, []byte(" "))
+		if found {
+			timestamp, err = time.Parse(time.RFC3339Nano, string(timestampPart))
+			if err != nil {
+				timestamp = time.Time{}
+			}
+			message = messagePart
+		}
+	}
+
+	entry := api.ContainerLogEntry{
+		Timestamp: timestamp,
+		// Clone is required because message is a slice into data, which stdcopy.StdCopy may reuse
+		// after Write returns but before the entry is consumed from the channel.
+		Message: bytes.Clone(message),
+	}
+	if w.isStderr {
+		entry.Stream = api.LogStreamStderr
+	} else {
+		entry.Stream = api.LogStreamStdout
+	}
+
+	select {
+	case w.ch <- entry:
+		return len(data), nil
+	case <-w.ctx.Done():
+		return 0, w.ctx.Err()
+	}
+}

internal/machine/firewall/iptables_darwin.go

@@ -1,9 +1,12 @@
 package firewall
 
-import "fmt"
+import (
+	"fmt"
+	"net/netip"
+)
 
 // ConfigureIptablesChains is a stub for Darwin.
-func ConfigureIptablesChains() error {
+func ConfigureIptablesChains(machineIP netip.Addr) error {
 	return fmt.Errorf("not supported on Darwin")
 }
 

internal/machine/firewall/iptables_linux.go

@@ -3,6 +3,7 @@ package firewall
 import (
 	"fmt"
 	"log/slog"
+	"net/netip"
 	"strconv"
 	"strings"
 
@@ -18,7 +19,7 @@ const (
 )
 
 // ConfigureIptablesChains sets up custom iptables chains and initial firewall rules for Uncloud networking.
-func ConfigureIptablesChains() error {
+func ConfigureIptablesChains(machineIP netip.Addr) error {
 	if err := createIptablesChains(); err != nil {
 		return err
 	}
@@ -27,10 +28,24 @@ func ConfigureIptablesChains() error {
 	ipt6 := iptables.GetIptable(iptables.IPv6)
 
 	// Allow WireGuard traffic to the machine.
-	acceptWireGuardRule := []string{"-p", "udp", "--dport", strconv.Itoa(network.WireGuardPort), "-j", "ACCEPT"}
-	err := ipt4.ProgramRule(iptables.Filter, UncloudInputChain, iptables.Insert, acceptWireGuardRule)
-	if err != nil {
-		return fmt.Errorf("insert iptables rule '%s': %w", strings.Join(acceptWireGuardRule, " "), err)
+	acceptWireGuardRule := []string{
+		"-p", "udp",
+		"--dport", strconv.Itoa(network.WireGuardPort),
+		"-j", "ACCEPT",
+	}
+	// Allow cluster machines to access the unregistry (embedded image registry) on the machine to push/pull images.
+	// TODO: allow access only from the machine IPs (10.210.N.1) but not the containers running on them. Use ipset?
+	acceptUnregistryRule := []string{
+		"-i", network.WireGuardInterfaceName,
+		"-d", machineIP.String(),
+		"-p", "tcp",
+		"--dport", strconv.Itoa(constants.UnregistryPort),
+		"-j", "ACCEPT",
+	}
+	for _, rule := range [][]string{acceptUnregistryRule, acceptWireGuardRule} {
+		if err := ipt4.ProgramRule(iptables.Filter, UncloudInputChain, iptables.Insert, rule); err != nil {
+			return fmt.Errorf("insert iptables rule '%s': %w", strings.Join(rule, " "), err)
+		}
 	}
 
 	// Allow cluster machines to access Machine API via the management IPv6 WireGuard network.
@@ -50,7 +65,7 @@ func ConfigureIptablesChains() error {
 		"-j", "ACCEPT",
 	}
 	for _, rule := range [][]string{acceptMachineAPIRule, acceptCorrosionGossipRule} {
-		if err = ipt6.ProgramRule(iptables.Filter, UncloudInputChain, iptables.Insert, rule); err != nil {
+		if err := ipt6.ProgramRule(iptables.Filter, UncloudInputChain, iptables.Insert, rule); err != nil {
 			return fmt.Errorf("insert ip6tables rule '%s': %w", strings.Join(rule, " "), err)
 		}
 	}

internal/machine/machine.go

@@ -30,6 +30,7 @@ import (
 	machinedocker "github.com/psviderski/uncloud/internal/machine/docker"
 	"github.com/psviderski/uncloud/internal/machine/network"
 	"github.com/psviderski/uncloud/internal/machine/store"
+	"github.com/psviderski/unregistry"
 	"github.com/siderolabs/grpc-proxy/proxy"
 	"golang.org/x/sync/errgroup"
 	"google.golang.org/grpc"
@@ -63,6 +64,8 @@ type Config struct {
 
 	// DockerClient manages system and user containers using the local Docker daemon.
 	DockerClient *client.Client
+	// ContainerdSockPath is the path to the containerd.sock used by Docker.
+	ContainerdSockPath string
 
 	// CaddyConfigDir specifies the directory where the machine generates the Caddy reverse proxy configuration file
 	// for routing external traffic to service containers across the internal network. Default is DataDir/caddy.
@@ -93,6 +96,26 @@ func (c *Config) SetDefaults() (*Config, error) {
 		}
 		cfg.DockerClient = cli
 	}
+	if cfg.ContainerdSockPath == "" {
+		// Auto-detect the containerd.sock path used by Docker.
+		paths := []string{
+			"/run/containerd/containerd.sock", // Default path on most Linux distributions.
+			"/run/docker/containerd/containerd.sock",
+			"/var/run/containerd/containerd.sock",
+			"/var/run/docker/containerd/containerd.sock",
+		}
+		for _, path := range paths {
+			if _, err := os.Stat(path); err == nil {
+				cfg.ContainerdSockPath = path
+				slog.Debug("Detected containerd socket used by Docker.", "path", path)
+				break
+			}
+		}
+
+		if cfg.ContainerdSockPath == "" {
+			slog.Warn("Failed to auto-detect containerd socket used by Docker.")
+		}
+	}
 
 	if cfg.CorrosionDir == "" {
 		cfg.CorrosionDir = filepath.Join(cfg.DataDir, "corrosion")
@@ -231,7 +254,6 @@ func NewMachine(config *Config) (*Machine, error) {
 	if err != nil {
 		return nil, fmt.Errorf("init machine database: %w", err)
 	}
-
 	dockerService := machinedocker.NewService(config.DockerClient, db)
 
 	// Init a local gRPC proxy server that proxies requests to the local or remote machine API servers.
@@ -260,9 +282,14 @@ func NewMachine(config *Config) (*Machine, error) {
 	internalDNSIP := func() netip.Addr {
 		return m.IP()
 	}
-	m.dockerServer = machinedocker.NewServer(dockerService, db, internalDNSIP,
-		machinedocker.WithNetworkReady(m.IsNetworkReady),
-		machinedocker.WithWaitForNetworkReady(m.WaitForNetworkReady))
+	// Machine ID will only be available after the machine is initialised as a cluster member so wrap it in a function.
+	machineID := func() string {
+		return m.state.ID
+	}
+	m.dockerServer = machinedocker.NewServer(dockerService, db, internalDNSIP, machineID, machinedocker.ServerOptions{
+		NetworkReady:        m.IsNetworkReady,
+		WaitForNetworkReady: m.WaitForNetworkReady,
+	})
 	caddyServer := caddyconfig.NewServer(caddyconfig.NewService(config.CaddyConfigDir))
 	m.localMachineServer = newGRPCServer(m, c, m.dockerServer, caddyServer)
 
@@ -403,11 +430,43 @@ func (m *Machine) Run(ctx context.Context) error {
 			}
 
 			dnsResolver := dns.NewClusterResolver(m.store)
-			dnsServer, err := dns.NewServer(m.IP(), dnsResolver, m.config.DNSUpstreams)
+			dnsServer, err := dns.NewServer(
+				m.IP(),
+				m.state.Network.Subnet,
+				dnsResolver,
+				m.config.DNSUpstreams,
+			)
 			if err != nil {
 				return fmt.Errorf("create embedded DNS server: %w", err)
 			}
 
+			var unreg *unregistry.Registry
+			if m.config.ContainerdSockPath != "" {
+				isContainerdStore, err := m.dockerService.IsContainerdImageStoreEnabled(ctx)
+				if err != nil {
+					return fmt.Errorf("check if Docker uses containerd image store: %w", err)
+				}
+
+				if isContainerdStore {
+					// Create an embedded container registry listening on the machine IP address and
+					// using the local Docker (containerd) image store as its backend.
+					unreg, err = unregistry.NewRegistry(unregistry.Config{
+						Addr:                net.JoinHostPort(m.IP().String(), strconv.Itoa(constants.UnregistryPort)),
+						ContainerdNamespace: "moby",
+						ContainerdSock:      m.config.ContainerdSockPath,
+						LogFormatter:        "text",
+						LogLevel:            "info",
+					})
+					if err != nil {
+						return fmt.Errorf("create embedded registry: %w", err)
+					}
+				} else {
+					slog.Warn("Skipping embedded unregistry setup as Docker is not using the containerd image store.")
+				}
+			} else {
+				slog.Warn("Skipping embedded unregistry setup as the containerd socket path is not configured.")
+			}
+
 			m.mu.Lock()
 			m.clusterCtrl, err = newClusterController(
 				m.state,
@@ -419,6 +478,7 @@ func (m *Machine) Run(ctx context.Context) error {
 				caddyconfigCtrl,
 				dnsServer,
 				dnsResolver,
+				unreg,
 			)
 			m.mu.Unlock()
 			if err != nil {
@@ -761,7 +821,14 @@ func (m *Machine) JoinCluster(_ context.Context, req *pb.JoinClusterRequest) (*e
 	if err := m.state.Save(); err != nil {
 		return nil, status.Errorf(codes.Internal, "save machine state: %v", err)
 	}
-	slog.Info("Machine configured to join the cluster.", "id", m.state.ID, "name", m.state.Name)
+	slog.Info(
+		"Machine configured to join the cluster.",
+		"id", m.state.ID,
+		"name", m.state.Name,
+		"subnet", m.state.Network.Subnet.String(),
+		"management_ip", m.state.Network.ManagementIP.String(),
+		"peers", len(m.state.Network.Peers),
+	)
 	// Signal that the machine is initialised as a member of a cluster.
 	m.initialised <- struct{}{}
 

internal/machine/network/wireguard_linux.go

@@ -79,7 +79,7 @@ func (n *WireGuardNetwork) Configure(config Config) error {
 	if err := n.configureDevice(config); err != nil {
 		return err
 	}
-	slog.Info("Configured WireGuard interface.", "name", n.link.Attrs().Name)
+	slog.Info("Configured WireGuard interface.", "name", n.link.Attrs().Name, "peers", len(n.peers))
 
 	managementPrefix, err := addrToSingleIPPrefix(config.ManagementIP)
 	if err != nil {

internal/machine/store/schema.sql

@@ -19,7 +19,7 @@ CREATE TABLE machines
 CREATE TABLE containers
 (
     id           TEXT NOT NULL PRIMARY KEY,
-    -- container is a JSON-serialized api.Container struct.
+    -- container is a JSON-serialized api.ServiceContainer struct.
     container    TEXT NOT NULL DEFAULT '{}' CHECK (json_valid(container)),
     machine_id   TEXT NOT NULL DEFAULT '',
     service_id   TEXT AS (json_extract(container, '$.Config.Labels."uncloud.service.id"')),

internal/proxy/proxy.go

@@ -0,0 +1,132 @@
+package proxy
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"net"
+	"os"
+	"sync"
+	"time"
+)
+
+// Proxy proxies local connections to a remote TCP address optionally using a custom dialer.
+type Proxy struct {
+	Listener    net.Listener
+	RemoteAddr  string
+	DialContext func(ctx context.Context, network, address string) (net.Conn, error)
+	OnError     func(error)
+	activeConns sync.WaitGroup
+}
+
+// deadliner is an interface for listeners that support setting deadlines.
+type deadliner interface {
+	SetDeadline(t time.Time) error
+}
+
+// halfCloser is an interface for connections that support half-close.
+type halfCloser interface {
+	CloseWrite() error
+}
+
+// Run starts the proxy and runs until the context is canceled.
+func (p *Proxy) Run(ctx context.Context) {
+	if p.DialContext == nil {
+		p.DialContext = (&net.Dialer{}).DialContext
+	}
+
+	defer p.Listener.Close()
+
+	// Handle incoming connections until context is canceled.
+Loop:
+	for {
+		select {
+		case <-ctx.Done():
+			break Loop
+		default:
+		}
+
+		// Set a deadline on the listener if supported to check context periodically.
+		if dl, ok := p.Listener.(deadliner); ok {
+			dl.SetDeadline(time.Now().Add(1 * time.Second))
+		}
+
+		conn, err := p.Listener.Accept()
+		if err != nil {
+			if os.IsTimeout(err) {
+				// Just a timeout, continue to check context and accept again.
+				continue
+			}
+
+			select {
+			case <-ctx.Done():
+				break Loop
+			default:
+				if p.OnError != nil {
+					p.OnError(fmt.Errorf("accept local connection: %w", err))
+				}
+				continue
+			}
+		}
+
+		p.activeConns.Add(1)
+		go p.handleConnection(ctx, conn)
+	}
+
+	// Wait for all connections to finish.
+	p.activeConns.Wait()
+}
+
+func (p *Proxy) handleConnection(ctx context.Context, localConn net.Conn) {
+	defer p.activeConns.Done()
+	defer localConn.Close()
+
+	// Use a separate context with timeout for dialing the remote address.
+	dialCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
+	defer cancel()
+
+	remoteConn, err := p.DialContext(dialCtx, "tcp", p.RemoteAddr)
+	if err != nil {
+		if p.OnError != nil {
+			p.OnError(fmt.Errorf("connect remote address '%s': %w", p.RemoteAddr, err))
+		}
+		return
+	}
+	defer remoteConn.Close()
+
+	// Bidirectional copy with proper half-close handling.
+	done := make(chan error, 2)
+
+	go func() {
+		_, err := io.Copy(remoteConn, localConn)
+		// Close write half of remote connection if supported.
+		if hc, ok := remoteConn.(halfCloser); ok {
+			hc.CloseWrite()
+		}
+		done <- err
+	}()
+
+	go func() {
+		_, err := io.Copy(localConn, remoteConn)
+		// Close write half of local connection if supported.
+		if hc, ok := localConn.(halfCloser); ok {
+			hc.CloseWrite()
+		}
+		done <- err
+	}()
+
+	// Wait for both copies to complete or context cancel.
+	for i := 0; i < 2; i++ {
+		select {
+		case <-ctx.Done():
+			// Close connections to abort ongoing copies.
+			localConn.Close()
+			remoteConn.Close()
+			return
+		case err = <-done:
+			if err != nil && p.OnError != nil {
+				p.OnError(fmt.Errorf("data copy: %w", err))
+			}
+		}
+	}
+}

internal/sshexec/sshcli.go

@@ -0,0 +1,72 @@
+package sshexec
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"os/exec"
+	"strconv"
+	"strings"
+)
+
+type SSHCLIRemote struct {
+	user    string
+	host    string
+	port    int
+	keyPath string
+}
+
+func NewSSHCLIRemote(user, host string, port int, keyPath string) *SSHCLIRemote {
+	return &SSHCLIRemote{
+		user:    user,
+		host:    host,
+		port:    port,
+		keyPath: keyPath,
+	}
+}
+
+// TODO: Refactor and reuse this with buildDialArgs and buildSSHArgs from
+// SSHCLI Connector.
+func (r *SSHCLIRemote) buildSSHArgs() []string {
+	args := []string{"-o", "ConnectTimeout=5"}
+
+	if r.port != 0 && r.port != 22 {
+		args = append(args, "-p", strconv.Itoa(r.port))
+	}
+
+	if r.keyPath != "" {
+		args = append(args, "-i", r.keyPath)
+	}
+
+	args = append(args, r.user+"@"+r.host)
+	return args
+}
+
+func (r *SSHCLIRemote) Run(ctx context.Context, cmd string) (string, error) {
+	args := r.buildSSHArgs()
+	args = append(args, cmd)
+
+	execCmd := exec.CommandContext(ctx, "ssh", args...)
+	output, err := execCmd.CombinedOutput()
+	if err != nil {
+		return strings.TrimSpace(string(output)),
+			fmt.Errorf("run command on remote host: %w: %s", err, string(output))
+	}
+	return strings.TrimSpace(string(output)), nil
+}
+
+func (r *SSHCLIRemote) Stream(ctx context.Context, cmd string, stdout, stderr io.Writer) error {
+	args := r.buildSSHArgs()
+	args = append(args, cmd)
+
+	execCmd := exec.CommandContext(ctx, "ssh", args...)
+	execCmd.Stdout = stdout
+	execCmd.Stderr = stderr
+
+	return execCmd.Run()
+}
+
+// no-op as there is no persistent connection.
+func (r *SSHCLIRemote) Close() error {
+	return nil
+}

internal/sshexec/sshcli_test.go

@@ -0,0 +1,76 @@
+package sshexec
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestSSHCLIRemote_buildSSHArgs(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name     string
+		user     string
+		host     string
+		port     int
+		keyPath  string
+		expected []string
+	}{
+		{
+			name:     "default port, no key",
+			user:     "root",
+			host:     "example.com",
+			port:     22,
+			keyPath:  "",
+			expected: []string{"-o", "ConnectTimeout=5", "root@example.com"},
+		},
+		{
+			name:     "custom port",
+			user:     "ubuntu",
+			host:     "192.168.1.10",
+			port:     2222,
+			keyPath:  "",
+			expected: []string{"-o", "ConnectTimeout=5", "-p", "2222", "ubuntu@192.168.1.10"},
+		},
+		{
+			name:     "with key path",
+			user:     "root",
+			host:     "example.com",
+			port:     22,
+			keyPath:  "/path/to/key",
+			expected: []string{"-o", "ConnectTimeout=5", "-i", "/path/to/key", "root@example.com"},
+		},
+		{
+			name:     "all options",
+			user:     "admin",
+			host:     "server.local",
+			port:     2222,
+			keyPath:  "~/.ssh/id_rsa",
+			expected: []string{"-o", "ConnectTimeout=5", "-p", "2222", "-i", "~/.ssh/id_rsa", "admin@server.local"},
+		},
+		{
+			name:     "port zero (default)",
+			user:     "root",
+			host:     "example.com",
+			port:     0,
+			keyPath:  "",
+			expected: []string{"-o", "ConnectTimeout=5", "root@example.com"},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			remote := &SSHCLIRemote{
+				user:    tt.user,
+				host:    tt.host,
+				port:    tt.port,
+				keyPath: tt.keyPath,
+			}
+			result := remote.buildSSHArgs()
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}

internal/ucind/cluster.go

@@ -8,10 +8,10 @@ import (
 	"time"
 
 	"github.com/cenkalti/backoff/v4"
+	"github.com/containerd/errdefs"
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/filters"
 	"github.com/docker/docker/api/types/network"
-	dockerclient "github.com/docker/docker/client"
 	"github.com/docker/go-connections/nat"
 	"github.com/psviderski/uncloud/internal/machine"
 	"github.com/psviderski/uncloud/internal/machine/api/pb"
@@ -142,7 +142,7 @@ func (p *Provisioner) initCluster(ctx context.Context, machines []Machine) error
 
 	// Join the rest of the machines to the cluster.
 	for _, m := range machines[1:] {
-		if err = WaitMachineReady(ctx, m, 5*time.Second); err != nil {
+		if err = WaitMachineReady(ctx, m, 30*time.Second); err != nil {
 			return fmt.Errorf("wait for machine %q to be ready: %w", m.Name, err)
 		}
 
@@ -200,7 +200,7 @@ func (p *Provisioner) InspectCluster(ctx context.Context, name string) (Cluster,
 	// Docker network name is the same as the cluster name.
 	net, err := p.dockerCli.NetworkInspect(ctx, name, network.InspectOptions{})
 	if err != nil {
-		if dockerclient.IsErrNotFound(err) {
+		if errdefs.IsNotFound(err) {
 			return c, ErrNotFound
 		}
 		return c, fmt.Errorf("inspect Docker network '%s': %w", name, err)

internal/ucind/machine.go

@@ -10,9 +10,9 @@ import (
 	"time"
 
 	"github.com/cenkalti/backoff/v4"
+	"github.com/containerd/errdefs"
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/image"
-	dockerclient "github.com/docker/docker/client"
 	"github.com/docker/go-connections/nat"
 	"github.com/psviderski/uncloud/internal/secret"
 	"github.com/psviderski/uncloud/pkg/client"
@@ -137,7 +137,7 @@ func (p *Provisioner) createContainerWithImagePull(
 		return resp, nil
 	}
 
-	if !dockerclient.IsErrNotFound(err) {
+	if !errdefs.IsNotFound(err) {
 		return resp, fmt.Errorf("create Docker container: %w", err)
 	}
 

misc/user_guide.md

@@ -22,7 +22,7 @@ Uncloud stores its configuration in `~/.config/uncloud/config.yaml`. If you wish
 
 Begin by initialising the first node in your cluster with `uc machine init [USER@HOST:PORT]`. If you do not have a need for Caddy reverse proxy, you may disable this feature with `--no-caddy`. If you want to avoid using uncloud's managed DNS service, add the `--no-dns` flag.
 
-This command will idempotently install Docker, uncloudd, uncloud-corrosion. If Caddy is enabled, it will set up a reverse proxy. If Uncloud DNS is enabled, it will create a DNS A record for the machine's public IP address under `*.[CLUSTER ID].cluster.uncloud.run`.
+This command will idempotently install Docker, uncloudd, uncloud-corrosion. If Caddy is enabled, it will set up a reverse proxy. If Uncloud DNS is enabled, it will create a DNS A record for the machine's public IP address under `*.[CLUSTER ID].uncld.dev`.
 
 If you wish to uninstall Uncloud and its components, run `uncloud-uninstall`.
 

pkg/api/client.go

@@ -2,13 +2,10 @@ package api
 
 import (
 	"context"
-	"fmt"
-	"slices"
 
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/volume"
 	"github.com/psviderski/uncloud/internal/machine/api/pb"
-	"google.golang.org/grpc/metadata"
 )
 
 type Client interface {
@@ -28,6 +25,7 @@ type ContainerClient interface {
 	RemoveContainer(ctx context.Context, serviceNameOrID, containerNameOrID string, opts container.RemoveOptions) error
 	StartContainer(ctx context.Context, serviceNameOrID, containerNameOrID string) error
 	StopContainer(ctx context.Context, serviceNameOrID, containerNameOrID string, opts container.StopOptions) error
+	ExecContainer(ctx context.Context, serviceNameOrID, containerNameOrID string, config ExecOptions) (int, error)
 }
 
 type DNSClient interface {
@@ -50,6 +48,8 @@ type ServiceClient interface {
 	RunService(ctx context.Context, spec ServiceSpec) (RunServiceResponse, error)
 	InspectService(ctx context.Context, id string) (Service, error)
 	RemoveService(ctx context.Context, id string) error
+	StopService(ctx context.Context, id string, opts container.StopOptions) error
+	StartService(ctx context.Context, id string) error
 }
 
 type VolumeClient interface {
@@ -57,27 +57,3 @@ type VolumeClient interface {
 	ListVolumes(ctx context.Context, filter *VolumeFilter) ([]MachineVolume, error)
 	RemoveVolume(ctx context.Context, machineNameOrID, volumeName string, force bool) error
 }
-
-// ProxyMachinesContext returns a new context that proxies gRPC requests to the specified machines.
-// If namesOrIDs is nil, all machines are included.
-func ProxyMachinesContext(
-	ctx context.Context, cli MachineClient, namesOrIDs []string,
-) (context.Context, MachineMembersList, error) {
-	machines, err := cli.ListMachines(ctx, nil)
-	if err != nil {
-		return nil, nil, fmt.Errorf("list machines: %w", err)
-	}
-
-	var proxiedMachines MachineMembersList
-	md := metadata.New(nil)
-	for _, m := range machines {
-		if len(namesOrIDs) == 0 ||
-			slices.Contains(namesOrIDs, m.Machine.Name) || slices.Contains(namesOrIDs, m.Machine.Id) {
-			proxiedMachines = append(proxiedMachines, m)
-			machineIP, _ := m.Machine.Network.ManagementIp.ToAddr()
-			md.Append("machines", machineIP.String())
-		}
-	}
-
-	return metadata.NewOutgoingContext(ctx, md), proxiedMachines, nil
-}

pkg/api/config.go

@@ -0,0 +1,125 @@
+// Implementation of Config feature from the Compose spec
+package api
+
+import (
+	"bytes"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strconv"
+)
+
+// ConfigSpec defines a configuration object that can be mounted into containers
+type ConfigSpec struct {
+	Name string
+
+	// Content of the config when specified inline
+	Content []byte `json:",omitempty"`
+
+	// Note: NOT IMPLEMENTED
+	// External indicates this config already exists and should not be created
+	// External bool `json:",omitempty"`
+
+	// Note: NOT IMPLEMENTED
+	// Labels for the config
+	// Labels map[string]string `json:",omitempty"`
+
+	// TODO: add support for "environment"
+}
+
+func (c *ConfigSpec) Validate() error {
+	if c.Name == "" {
+		return fmt.Errorf("config name is required")
+	}
+	return nil
+}
+
+// Equals compares two ConfigSpec instances
+func (c *ConfigSpec) Equals(other ConfigSpec) bool {
+	return c.Name == other.Name &&
+		bytes.Equal(c.Content, other.Content)
+}
+
+// ConfigMount defines how a config is mounted into a container
+type ConfigMount struct {
+	// ConfigName references a config defined in ServiceSpec.Configs by its Name field
+	ConfigName string
+	// ContainerPath is the absolute path where the config is mounted in the container
+	ContainerPath string `json:",omitempty"`
+	// Uid for the mounted config file
+	Uid string `json:",omitempty"`
+	// Gid for the mounted config file
+	Gid string `json:",omitempty"`
+	// Mode (file permissions) for the mounted config file
+	Mode *os.FileMode `json:",omitempty"`
+}
+
+func (c *ConfigMount) GetNumericUid() (*uint64, error) {
+	if c.Uid == "" {
+		return nil, nil
+	}
+	uid, err := strconv.ParseUint(c.Uid, 10, 64)
+	if err != nil {
+		return nil, fmt.Errorf("invalid Uid '%s': %w", c.Uid, err)
+	}
+	if int(uid) < 0 {
+		return nil, fmt.Errorf("invalid Uid '%s': value too high", c.Uid)
+	}
+	return &uid, nil
+}
+
+func (c *ConfigMount) GetNumericGid() (*uint64, error) {
+	if c.Gid == "" {
+		return nil, nil
+	}
+	gid, err := strconv.ParseUint(c.Gid, 10, 64)
+	if err != nil {
+		return nil, fmt.Errorf("invalid Gid '%s': %w", c.Gid, err)
+	}
+	if int(gid) < 0 {
+		return nil, fmt.Errorf("invalid Gid '%s': value too high", c.Gid)
+	}
+	return &gid, nil
+}
+
+func (c *ConfigMount) Validate() error {
+	if c.ConfigName == "" {
+		return fmt.Errorf("config mount source is required")
+	}
+	if _, err := c.GetNumericUid(); err != nil {
+		return err
+	}
+	if _, err := c.GetNumericGid(); err != nil {
+		return err
+	}
+	if c.ContainerPath != "" && !filepath.IsAbs(c.ContainerPath) {
+		return fmt.Errorf("container path must be absolute")
+	}
+	return nil
+}
+
+// ValidateConfigsAndMounts takes config specs and config mounts and validates that all mounts refer to existing specs
+func ValidateConfigsAndMounts(configs []ConfigSpec, mounts []ConfigMount) error {
+	configMap := make(map[string]struct{})
+	for _, cfg := range configs {
+		if err := cfg.Validate(); err != nil {
+			return fmt.Errorf("invalid config: %w", err)
+		}
+		if _, ok := configMap[cfg.Name]; ok {
+			return fmt.Errorf("duplicate config name: '%s'", cfg.Name)
+		}
+
+		configMap[cfg.Name] = struct{}{}
+	}
+
+	for _, mount := range mounts {
+		if err := mount.Validate(); err != nil {
+			return fmt.Errorf("invalid config mount: %w", err)
+		}
+		if _, exists := configMap[mount.ConfigName]; !exists {
+			return fmt.Errorf("config mount source '%s' does not refer to any defined config", mount.ConfigName)
+		}
+	}
+
+	return nil
+}

pkg/api/config_test.go

@@ -0,0 +1,294 @@
+package api
+
+import (
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// uint64Ptr is a convenience function to create a pointer to a uint64 value
+func uint64Ptr(v uint64) *uint64 {
+	return &v
+}
+
+func TestConfigMount_GetNumericUid(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name     string
+		uid      string
+		expected *uint64
+		wantErr  string
+	}{
+		{
+			name:     "empty uid returns nil",
+			uid:      "",
+			expected: nil,
+		},
+		{
+			name:     "valid numeric uid",
+			uid:      "1000",
+			expected: uint64Ptr(1000),
+		},
+		{
+			name:     "zero uid",
+			uid:      "0",
+			expected: uint64Ptr(0),
+		},
+		{
+			name:    "invalid non-numeric uid",
+			uid:     "root",
+			wantErr: "invalid Uid 'root'",
+		},
+		{
+			name:    "negative uid",
+			uid:     "-1",
+			wantErr: "invalid Uid",
+		},
+		{
+			name:    "very large uid",
+			uid:     "18446744073709551615", // max uint64
+			wantErr: "invalid Uid '18446744073709551615': value too high",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			mount := &ConfigMount{Uid: tt.uid}
+			uid, err := mount.GetNumericUid()
+
+			if tt.wantErr != "" {
+				require.Error(t, err)
+				assert.Contains(t, err.Error(), tt.wantErr)
+				assert.Nil(t, uid)
+				return
+			}
+
+			require.NoError(t, err)
+			if tt.expected == nil {
+				assert.Nil(t, uid)
+			} else {
+				require.NotNil(t, uid)
+				assert.Equal(t, *tt.expected, *uid)
+			}
+		})
+	}
+}
+
+func TestConfigMount_GetNumericGid(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name     string
+		gid      string
+		expected *uint64
+		wantErr  string
+	}{
+		{
+			name:     "empty gid returns nil",
+			gid:      "",
+			expected: nil,
+		},
+		{
+			name:     "valid numeric gid",
+			gid:      "1000",
+			expected: uint64Ptr(1000),
+		},
+		{
+			name:     "zero gid",
+			gid:      "0",
+			expected: uint64Ptr(0),
+		},
+		{
+			name:    "invalid non-numeric gid",
+			gid:     "wheel",
+			wantErr: "invalid Gid 'wheel'",
+		},
+		{
+			name:    "negative gid",
+			gid:     "-1",
+			wantErr: "invalid Gid",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			mount := &ConfigMount{Gid: tt.gid}
+			gid, err := mount.GetNumericGid()
+
+			if tt.wantErr != "" {
+				require.Error(t, err)
+				assert.Contains(t, err.Error(), tt.wantErr)
+				assert.Nil(t, gid)
+				return
+			}
+
+			require.NoError(t, err)
+			if tt.expected == nil {
+				assert.Nil(t, gid)
+			} else {
+				require.NotNil(t, gid)
+				assert.Equal(t, *tt.expected, *gid)
+			}
+		})
+	}
+}
+
+func TestValidateConfigsAndMounts(t *testing.T) {
+	t.Parallel()
+
+	mode := os.FileMode(0o644)
+
+	tests := []struct {
+		name    string
+		configs []ConfigSpec
+		mounts  []ConfigMount
+		wantErr string
+	}{
+		{
+			name:    "empty configs and mounts",
+			configs: []ConfigSpec{},
+			mounts:  []ConfigMount{},
+		},
+		{
+			name: "valid configs without mounts",
+			configs: []ConfigSpec{
+				{Name: "config1", Content: []byte("content1")},
+				{Name: "config2", Content: []byte("content2")},
+			},
+			mounts: []ConfigMount{},
+		},
+		{
+			name: "valid configs with valid mounts",
+			configs: []ConfigSpec{
+				{Name: "config1", Content: []byte("content1")},
+				{Name: "config2", Content: []byte("content2")},
+			},
+			mounts: []ConfigMount{
+				{ConfigName: "config1", ContainerPath: "/etc/config1"},
+				{ConfigName: "config2", ContainerPath: "/etc/config2", Uid: "1000", Gid: "1000"},
+			},
+		},
+		{
+			name: "config with empty name",
+			configs: []ConfigSpec{
+				{Name: "", Content: []byte("content")},
+			},
+			mounts:  []ConfigMount{},
+			wantErr: "config name is required",
+		},
+		{
+			name: "duplicate config names",
+			configs: []ConfigSpec{
+				{Name: "config1", Content: []byte("content1")},
+				{Name: "config1", Content: []byte("content2")},
+			},
+			mounts:  []ConfigMount{},
+			wantErr: "duplicate config name: 'config1'",
+		},
+		{
+			name: "mount with empty config name",
+			configs: []ConfigSpec{
+				{Name: "config1", Content: []byte("content1")},
+			},
+			mounts: []ConfigMount{
+				{ConfigName: "", ContainerPath: "/etc/config"},
+			},
+			wantErr: "config mount source is required",
+		},
+		{
+			name: "mount referencing non-existent config",
+			configs: []ConfigSpec{
+				{Name: "config1", Content: []byte("content1")},
+			},
+			mounts: []ConfigMount{
+				{ConfigName: "nonexistent", ContainerPath: "/etc/config"},
+			},
+			wantErr: "config mount source 'nonexistent' does not refer to any defined config",
+		},
+		{
+			name: "mount with invalid uid",
+			configs: []ConfigSpec{
+				{Name: "config1", Content: []byte("content1")},
+			},
+			mounts: []ConfigMount{
+				{ConfigName: "config1", ContainerPath: "/etc/config", Uid: "invalid"},
+			},
+			wantErr: "invalid Uid 'invalid'",
+		},
+		{
+			name: "mount with invalid gid",
+			configs: []ConfigSpec{
+				{Name: "config1", Content: []byte("content1")},
+			},
+			mounts: []ConfigMount{
+				{ConfigName: "config1", ContainerPath: "/etc/config", Gid: "invalid"},
+			},
+			wantErr: "invalid Gid 'invalid'",
+		},
+		{
+			name: "mount with relative container path",
+			configs: []ConfigSpec{
+				{Name: "config1", Content: []byte("content1")},
+			},
+			mounts: []ConfigMount{
+				{ConfigName: "config1", ContainerPath: "relative/path"},
+			},
+			wantErr: "container path must be absolute",
+		},
+		{
+			name: "mount with empty container path",
+			configs: []ConfigSpec{
+				{Name: "config1", Content: []byte("content1")},
+			},
+			mounts: []ConfigMount{
+				{ConfigName: "config1", ContainerPath: ""},
+			},
+			// Empty path is allowed
+		},
+		{
+			name: "mount with absolute container path",
+			configs: []ConfigSpec{
+				{Name: "config1", Content: []byte("content1")},
+			},
+			mounts: []ConfigMount{
+				{ConfigName: "config1", ContainerPath: "/absolute/path"},
+			},
+		},
+		{
+			name: "complex valid scenario",
+			configs: []ConfigSpec{
+				{Name: "nginx-conf", Content: []byte("server { listen 80; }")},
+				{Name: "app-config", Content: []byte("debug=true")},
+				{Name: "cert", Content: []byte("-----BEGIN CERTIFICATE-----")},
+			},
+			mounts: []ConfigMount{
+				{ConfigName: "nginx-conf", ContainerPath: "/etc/nginx/nginx.conf", Uid: "0", Gid: "0", Mode: &mode},
+				{ConfigName: "app-config", ContainerPath: "/app/config.env"},
+				{ConfigName: "cert", ContainerPath: "/etc/ssl/cert.pem", Uid: "1000", Gid: "1000"},
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			err := ValidateConfigsAndMounts(tt.configs, tt.mounts)
+
+			if tt.wantErr != "" {
+				require.Error(t, err)
+				assert.Contains(t, err.Error(), tt.wantErr)
+				return
+			}
+
+			require.NoError(t, err)
+		})
+	}
+}

pkg/api/container.go

@@ -7,7 +7,8 @@ import (
 	"strings"
 	"time"
 
-	"github.com/docker/docker/api/types"
+	"github.com/docker/docker/api/types/container"
+	"github.com/docker/docker/pkg/stringid"
 	"github.com/docker/go-units"
 )
 
@@ -23,7 +24,7 @@ const (
 )
 
 type Container struct {
-	types.ContainerJSON
+	container.InspectResponse
 	// created caches the parsed creation time by CreatedTime.
 	created time.Time
 }
@@ -52,7 +53,7 @@ func (c *Container) Healthy() bool {
 		return true
 	}
 
-	return c.State.Health.Status == types.Healthy
+	return c.State.Health.Status == container.Healthy
 }
 
 // HumanState returns a human-readable description of the container's state. Based on the Docker implementation:
@@ -78,7 +79,7 @@ func (c *Container) HumanState() (string, error) {
 
 		if c.State.Health != nil {
 			status := c.State.Health.Status
-			if status == types.Starting {
+			if status == container.Starting {
 				status = "health: " + status
 			}
 
@@ -144,6 +145,11 @@ type ServiceContainer struct {
 	ServiceSpec ServiceSpec
 }
 
+// ShortID returns the truncated ID of the container (12 characters).
+func (c *ServiceContainer) ShortID() string {
+	return stringid.TruncateID(c.ID)
+}
+
 // ServiceID returns the ID of the service this container belongs to.
 func (c *ServiceContainer) ServiceID() string {
 	return c.Config.Labels[LabelServiceID]