Install guide and config at curatedmcp.com

Docker MCP: Control Your Entire Docker Environment Through Natural Language

Managing Docker from the command line is powerful, but it's context-switching hell when you're deep in development or debugging. Docker MCP bridges that gap by giving AI agents like Claude, Cursor, and Windsurf direct access to your Docker environment—so you can orchestrate containers, images, volumes, and compose stacks without leaving your editor or chat window.

Built and maintained by Docker itself, this official MCP server connects to your local Docker socket. No remote credentials, no API keys, no security theater. Just native Docker control through natural conversation.

What It Does

Docker MCP unlocks container management as a first-class AI capability. Ask Claude to start your entire stack, inspect resource usage, stream logs, or build images—and it executes directly against your Docker daemon.

The server handles the full Docker lifecycle: list and filter containers, start/stop/remove them, build images from Dockerfiles, manage volumes and networks, run docker-compose commands, and even execute commands inside running containers. You get real-time log streaming, health inspection, and resource monitoring. Push and pull from Docker Hub or private registries without manually authenticating.

This means your AI agent can reason about your infrastructure, debug deployment issues, provision local environments, and iterate on container configs—all in the context of your current task.

How to Install

npx -y @docker/mcp-server

Add to your Claude Desktop config:

{
  "mcpServers": {
    "docker-mcp": {
      "command": "npx -y @docker/mcp-server"
    }
  }
}

Same approach works for Cursor and Windsurf. Restart your AI agent and you're live.

Real-World Use Cases

• Local development loops: "Start postgres and redis from my compose file, then run my migrations" — orchestrate your entire dev environment in one natural-language step instead of juggling multiple terminals.

• Debugging in production-like environments: "Show me the last 10 minutes of logs from the api container and tell me if we're hitting memory limits" — diagnose issues without SSH sessions or external monitoring dashboards.

• CI/CD-like workflows locally: "Build a new image from the current directory, tag it v2.1, and push it to my private registry" — test deployment automation locally before committing.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.