cgx-mcp-server-docker 0.3.0

A Docker MCP Server with enhanced session management and streamable-http support

🐋 Docker MCP server

An MCP server for managing Docker with natural language!

🔧 FIXED: Resolved anyio.ClosedResourceError and added streamable-http transport support!

🪩 What can it do?

• 🚀 Compose containers with natural language
• 🔍 Introspect & debug running containers
• 📀 Manage persistent data with Docker volumes
• 🌐 Works with both stdio and streamable-http transports
• 🔧 Enhanced error handling and session management

❓ Who is this for?

• Server administrators: connect to remote Docker engines for e.g. managing a public-facing website.
• Tinkerers: run containers locally and experiment with open-source apps supporting Docker.
• AI enthusiasts: push the limits of that an LLM is capable of!
• Agno app users: seamless integration with improved stability

🆕 Recent Improvements

• Fixed anyio.ClosedResourceError: Improved session management prevents premature session closure
• Streamable HTTP Support: Added proper HTTP transport for better integration with modern MCP clients
• Enhanced Error Handling: Comprehensive error handling prevents crashes and provides better debugging
• Agno Integration: Specifically tested and optimized for Agno app compatibility
• Health Monitoring: Added health check endpoint for HTTP transport monitoring

Demo

A quick demo showing a WordPress deployment using natural language:

https://github.com/user-attachments/assets/65e35e67-bce0-4449-af7e-9f4dd773b4b3

🏎️ Quickstart

Transport Options

Streamable HTTP (Recommended for Agno)

# Install dependencies
pip install -e .

# Start server with HTTP transport
python run_server.py --transport streamable-http --host 0.0.0.0 --port 8080

# Test health endpoint
curl http://localhost:8080/health

Standard I/O (Claude Desktop)

# Install dependencies
pip install -e .

# Start server with stdio transport
python run_server.py --transport stdio

Install

Claude Desktop

On MacOS: ~/Library/Application\ Support/Claude/claude_desktop_config.json

On Windows: %APPDATA%/Claude/claude_desktop_config.json

Install from PyPi with uv

If you don't have uv installed, follow the installation instructions for your system: link

Then add the following to your MCP servers file:

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

Install with Docker

Purely for convenience, the server can run in a Docker container.

After cloning this repository, build the Docker image:

docker build -t mcp-server-docker .

And then add the following to your MCP servers file:

"mcpServers": {
  "mcp-server-docker": {
    "command": "docker",
    "args": [
      "run",
      "-i",
      "--rm",
      "-v",
      "/var/run/docker.sock:/var/run/docker.sock",
      "mcp-server-docker:latest"
    ]
  }
}

Note that we mount the Docker socket as a volume; this ensures the MCP server can connect to and control the local Docker daemon.

📝 Prompts

🎻 docker_compose

Use natural language to compose containers. See above for a demo.

Provide a Project Name, and a description of desired containers, and let the LLM do the rest.

This prompt instructs the LLM to enter a plan+apply loop. Your interaction with the LLM will involve the following steps:

1. You give the LLM instructions for which containers to bring up
2. The LLM calculates a concise natural language plan and presents it to you
3. You either:
   • Apply the plan
   • Provide the LLM feedback, and the LLM recalculates the plan

Examples

• name: nginx, containers: "deploy an nginx container exposing it on port 9000"
• name: wordpress, containers: "deploy a WordPress container and a supporting MySQL container, exposing Wordpress on port 9000"

Resuming a Project

When starting a new chat with this prompt, the LLM will receive the status of any containers, volumes, and networks created with the given project name.

This is mainly useful for cleaning up, in-case you lose a chat that was responsible for many containers.

📔 Resources

The server implements a couple resources for every container:

• Stats: CPU, memory, etc. for a container
• Logs: tail some logs from a container

🔨 Tools

Containers

• list_containers
• create_container
• run_container
• recreate_container
• start_container
• fetch_container_logs
• stop_container
• remove_container

Images

• list_images
• pull_image
• push_image
• build_image
• remove_image

Networks

• list_networks
• create_network
• remove_network

Volumes

• list_volumes
• create_volume
• remove_volume

🚧 Disclaimers

Sensitive Data

DO NOT CONFIGURE CONTAINERS WITH SENSITIVE DATA. This includes API keys, database passwords, etc.

Any sensitive data exchanged with the LLM is inherently compromised, unless the LLM is running on your local machine.

If you are interested in securely passing secrets to containers, file an issue on this repository with your use-case.

Reviewing Created Containers

Be careful to review the containers that the LLM creates. Docker is not a secure sandbox, and therefore the MCP server can potentially impact the host machine through Docker.

For safety reasons, this MCP server doesn't support sensitive Docker options like --privileged or --cap-add/--cap-drop. If these features are of interest to you, file an issue on this repository with your use-case.

🛠️ Configuration

This server uses the Python Docker SDK's from_env method. For configuration details, see the documentation.

Connect to Docker over SSH

This MCP server can connect to a remote Docker daemon over SSH.

Simply set a ssh:// host URL in the MCP server definition:

"mcpServers": {
  "mcp-server-docker": {
    "command": "uvx",
    "args": [
      "mcp-server-docker"
    ],
    "env": {
      "DOCKER_HOST": "ssh://[e-mail]"
    }
  }
}

💻 Development

Prefer using Devbox to configure your development environment.

See the devbox.json for helpful development commands.

After setting up devbox you can configure your Claude MCP config to use it:

  "docker": {
    "command": "/path/to/repo/.devbox/nix/profile/default/bin/uv",
    "args": [
      "--directory",
      "/path/to/repo/",
      "run",
      "mcp-server-docker"
    ]
  },