docker-mcp-server 0.1.0

Docker MCP server exposing Docker operations as MCP tools

Docker MCP Server

A Model Context Protocol (MCP) server that exposes Docker operations as tools — manage containers, images, compose services, volumes, and networks directly from Claude.

Installation

pip install docker-mcp

Or with uvx (no install needed):

uvx docker-mcp

MCP Configuration

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "docker": {
      "command": "uvx",
      "args": ["docker-mcp"]
    }
  }
}

Claude Code

claude mcp add docker -- uvx docker-mcp

Or add manually to your project's .claude/settings.json:

{
  "mcpServers": {
    "docker": {
      "command": "uvx",
      "args": ["docker-mcp"]
    }
  }
}

Docker (SSE mode)

Run the server as a container and connect over HTTP:

docker run -d \
  -p 8000:8000 \
  -v /var/run/docker.sock:/var/run/docker.sock \
  ghcr.io/firasmosbahi/docker-mcp:latest

Then configure your MCP client to connect to http://localhost:8000/sse.

Available Tools

Containers

Tool	Description
list_containers	List containers (all=true includes stopped)
create_container	Create a container without starting it
run_container	Create and start a container in one step
start_container	Start a stopped container
stop_container	Stop a running container
restart_container	Restart a container
remove_container	Remove a container (force=true for running)
get_container_logs	Get container logs (configurable tail lines)
exec_in_container	Execute a command inside a running container
inspect_container	Get full container details as JSON
container_stats	Get CPU and memory usage for a container

Examples:

List all containers including stopped ones
→ list_containers(all=true)

Run an nginx container on port 8080
→ run_container(image="nginx", name="my-nginx", ports={"80/tcp": 8080}, detach=true)

Get the last 50 lines of logs from my-app
→ get_container_logs(container_id="my-app", tail=50)

Execute a command inside a container
→ exec_in_container(container_id="my-app", command="env")

Images

Tool	Description
list_images	List all local images with size
pull_image	Pull an image from a registry
build_image	Build an image from a Dockerfile
tag_image	Tag an image with a new name
remove_image	Remove an image
inspect_image	Get full image details as JSON

Examples:

Pull the latest Alpine image
→ pull_image(image="alpine", tag="latest")

Build an image from the current directory
→ build_image(path="/my/project", tag="my-app:v1")

Tag an image for pushing to a registry
→ tag_image(image="my-app:v1", repository="myrepo/my-app", tag="v1.0.0")

Compose

Tool	Description
compose_up	Start services (detached by default)
compose_down	Stop and remove services
compose_build	Build or rebuild services
compose_restart	Restart services
compose_ps	List service status
compose_logs	Get service logs

Examples:

Start all services in a project
→ compose_up(project_directory="/my/project")

Rebuild a specific service without cache
→ compose_build(project_directory="/my/project", service="api", no_cache=true)

Get logs for the web service
→ compose_logs(project_directory="/my/project", service="web", tail=100)

Volumes

Tool	Description
list_volumes	List all volumes
create_volume	Create a named volume
remove_volume	Remove a volume
inspect_volume	Get volume details

Networks

Tool	Description
list_networks	List all networks
create_network	Create a network
remove_network	Remove a network
inspect_network	Get network details
connect_container_to_network	Connect a container to a network
disconnect_container_from_network	Disconnect a container from a network

System

Tool	Description
system_info	Docker version, OS, container/image counts, disk usage
prune_containers	Remove all stopped containers
prune_images	Remove unused (dangling) images
prune_volumes	Remove unused volumes
prune_networks	Remove unused networks

Examples:

Check Docker system status and disk usage
→ system_info()

Free up space by removing unused resources
→ prune_containers()
→ prune_images()
→ prune_volumes()

Transport Modes

# stdio (default) — for MCP clients that manage the process
docker-mcp

# SSE — for HTTP-based MCP clients
docker-mcp --transport sse --port 8000

Requirements

• Python 3.11+
• Docker daemon running and accessible (Docker Desktop or Docker Engine)

Development

git clone https://github.com/firasmosbahi/docker-mcp
cd docker-mcp
poetry install
poetry run pytest