MCP Docker Server Documentation

A Model Context Protocol (MCP) server that exposes Docker functionality to AI assistants like Claude. Manage containers, images, networks, and volumes through a type-safe, documented API with safety controls.

Features

36 Docker Tools: Complete container, image, network, volume, and system management
5 AI Prompts: Intelligent troubleshooting, optimization, networking debug, and security analysis
2 Resources: Real-time container logs and resource statistics
Type Safety: Full type hints with Pydantic validation and mypy strict mode
Safety Controls: Three-tier safety system (safe/moderate/destructive) with configurable restrictions
Comprehensive Testing: Extensive test coverage with unit and integration tests
Modern Python: Built with Python 3.11+, uv package manager, and async-first design

Canonical Docs

The project docs live in three Markdown sources. Each section in this site links directly to them:

README.md — Overview, features, install flow, tools/prompts/resources
CONFIGURATION.md — All environment variables for Docker, transports, safety, and server settings
SECURITY.md — Threat model, OAuth/TLS guidance, deployment checklist

Quick Start

See Installation Instructions in the main README for a complete setup guide including:

Prerequisites (Python 3.11+, Docker, uv/pip)
Installation methods (uvx, uv, pip)
Configuration for Claude Desktop (Linux/macOS/Windows)
Platform-specific Docker socket URLs

Safety System

Three-tier classification: SAFE (read-only) → MODERATE (create/modify) → DESTRUCTIVE (delete).

Control via environment variables: SAFETY_ALLOW_DESTRUCTIVE_OPERATIONS, SAFETY_ALLOW_PRIVILEGED_CONTAINERS, etc.

See README.md for complete safety system documentation.

What’s Available

36 Docker Tools - Container, image, network, volume, and system management
5 AI Prompts - Troubleshooting, optimization, networking debug, security audit, compose generation
2 Resources - Container logs and stats streaming

For complete list see README.md.

Example Usage

List All Containers

# Using the MCP tool
result = await client.call_tool("docker_list_containers", {
    "all": True,
    "filters": {"status": ["running"]}
})

Troubleshoot a Container

# Using the AI prompt
prompt = await client.get_prompt("troubleshoot_container", {
    "container_id": "my-container"
})
# Returns detailed diagnostics and recommendations

Stream Container Logs

# Using the resource
logs = await client.read_resource("container://logs/my-container")
# Pass tail/follow arguments via your MCP client options

Documentation Sources

README.md: Overview, onboarding, tool/prompt/resource catalog
CONFIGURATION.md: Canonical environment variable reference (Docker, transports, safety, TLS)
SECURITY.md: OAuth/TLS configuration, IP filtering, audit trail, production hardening guide

For questions or issues, open a ticket in the GitHub repository.

Contributing

Contributions are welcome! Please see the GitHub repository for:

Filing issues and bug reports
Submitting pull requests
Reviewing code and documentation

License

MIT License - see LICENSE for details.

Support

Issues: GitHub Issues
Discussions: GitHub Discussions
MCP Documentation: modelcontextprotocol.io

Version: 1.1.0 Last Updated: 2025-11-14 Python: 3.11+ Docker: API version 1.41+