Infrastructure★★★★4/5
DockerMCPServer
The Docker MCP server provides 17 tools covering the full container lifecycle: create, start, stop, remove containers; pull and manage images; configure networks and volumes. The standout feature is the docker_compose prompt that guides the LLM through a plan+apply loop — describe what you want deployed, review the plan, then execute.
Remote Docker host support via SSH means you can manage containers on remote servers from your local AI assistant. Useful for homelab management without opening your Docker socket to the network.
Requires Python/uvx runtime. Note the GPL-3.0 license if that matters for your use case.
Pros
- + 17 tools covering containers, images, networks, and volumes
- + Plan+apply loop prevents blind execution
- + Remote Docker host support via SSH
- + Natural language container deployment
Cons
- - GPL-3.0 license (more restrictive than MIT)
- - Requires Python/uvx runtime
- - Direct Docker socket access — security implications
- - No Docker Compose file support (manages individual containers)
How We Use It
The nxsi stack is 7 containers (PostgreSQL, Redis, n8n, n8n-worker, API, dashboard, Caddy) plus monitoring (Prometheus, Grafana, Loki, Promtail). Docker MCP handles the routine container operations — checking which services are healthy, pulling updated images, restarting containers after config changes.
The remote SSH support is the key feature for us. Claude Code runs on a local machine but the Docker stack lives on a separate server. Being able to inspect container logs, check resource usage, and restart services without manually SSH-ing in and running commands saves context switches. The plan+apply loop adds a safety net — it shows what it's about to do before executing, which matters when you're touching production containers.
dockercontainersinfrastructureself-hosteddevops