MCP Docker Server

MCP Docker Server

A Python-based MCP (Model Context Protocol) server that provides secure Docker command execution for language models or other clients running in isolated environments like containers.

Overview

This MCP server runs on a host system and provides access to its Docker daemon via the MCP protocol. It implements security filtering and command validation to ensure safe operation, only allowing docker and docker-compose commands to be executed.

Features

• Docker Command Execution: Executes docker and docker-compose commands on the host.
• Docker Compose Support: Handles both legacy docker-compose and modern docker compose syntax.
• Security Validation: An explicit allowlist restricts executable commands to docker and docker-compose only.
• Async I/O: Built with asyncio for non-blocking command execution.
• Multiple Transports: Supports stdio, sse, and streamable-http MCP transports.
• Rich Toolset: Provides specific tools for common operations like listing containers, images, and checking system info.
• Error Handling: Returns detailed error messages for failed commands.

Architecture

Installation

1. Python: Requires Python 3.12 or newer.
2. Docker: Docker must be installed and the daemon must be running on the host system.
3. Dependencies: Install the required Python packages using uv or pip from pyproject.toml.
   # Using uv uv pip install -r requirements.txt # Or directly from the pyproject.toml uv pip install .

Usage

Starting the Server

The server can be started with different MCP transports.

Environment Variables

Configure the server's network binding with these environment variables.

• FASTMCP_HOST: Server bind address (default: 0.0.0.0 for container access).
• FASTMCP_PORT: Server port (default: 3000).

You can create a .env file to manage these variables:

Then run the server with:

MCP Protocol

The server uses the mcp library to expose tools. The available transports are stdio, sse, and streamable-http.

Available Methods (Tools)

The following tools are exposed by the server:

• execute_docker_command(command: str, working_directory: str = None) -> str
  • Description: Executes a general Docker or Docker Compose command. The command is validated against an allowlist.
  • Example: session.call_tool("execute_docker_command", {"command": "docker ps -a"})
• docker_system_info() -> str
  • Description: Retrieves Docker version and system disk usage information.
  • Example: session.call_tool("docker_system_info", {})
• list_containers(all: bool = False) -> str
  • Description: Lists Docker containers.
  • Args: all (boolean) - If true, shows all containers (including stopped ones). Defaults to False.
  • Example: session.call_tool("list_containers", {"all": True})
• list_images(all: bool = False) -> str
  • Description: Lists Docker images.
  • Args: all (boolean) - If true, shows all images (including intermediate ones). Defaults to False.
  • Example: session.call_tool("list_images", {})
• docker_compose_status(working_directory: str = None) -> str
  • Description: Gets the status of services defined in a docker-compose.yml file.
  • Args: working_directory (string) - The path to the directory containing the compose file.
  • Example: session.call_tool("docker_compose_status", {"working_directory": "/path/to/project"})

Security Features

Command Validation

The server uses an allowlist-based approach for security. The is_allowed_command method in main.py ensures that only commands beginning with docker or docker-compose (including docker compose) are processed. All other commands are rejected.

Error Handling

If a command fails to execute, the server captures stdout, stderr, and the exit code, returning a detailed error message to the client. The execute_command function contains a try...except block to handle exceptions during subprocess execution.

Troubleshooting

Common Issues

1. Docker Not Available:
   • Ensure Docker is installed: docker --version
   • Check that the Docker daemon is running: docker info
   • Verify your user has permissions to access the Docker socket.
2. Port Already in Use:
   • If you see an error like address already in use, the port (default 3000) is occupied.
   • Change the port using the FASTMCP_PORT environment variable.
   • Find the process using the port: lsof -i :3000 or netstat -tulnp | grep 3000.
3. Permission Denied (Docker Socket):
   • If you get a "permission denied" error when running Docker commands, add your user to the docker group: sudo usermod -aG docker $USER.
   • You will need to start a new shell session for this change to take effect.
4. Connection Refused:
   • Verify the MCP server is running and listening on the correct host and port.
   • Check for firewall rules that might be blocking the connection.
   • Ensure your client configuration matches the server's FASTMCP_HOST and FASTMCP_PORT.

Development

Project Structure

Testing

Run the example test client to connect to a running server, list its tools, and execute a sample command.

License

This project is open-source and available for modification and distribution.

Support

For issues and questions:

1. Check the Troubleshooting section.
2. Verify your Docker installation and permissions.
3. Review the server logs for error details.
4. Test connectivity with simple commands first (e.g., docker_system_info).