Skip to content
All experiments

Local DevOps Assistant

An MCP server that lets an AI assistant inspect and diagnose local Docker development environments through structured tools.

In progressStill being tested

Hypothesis

Learn the Model Context Protocol from first principles by building something real with it — a server that lets an MCP-compatible AI assistant inspect a local Docker environment — documented stage by stage rather than copied from a template.

Outcome

Six tools shipped, all read-only by design (list/inspect containers, tail logs, check a port, confirm the Docker daemon itself is responsive) — no tool ever builds a shell command from AI-provided input. Built on the official mcp Python SDK over stdio, deliberately without LangChain, RAG or any vector DB, so the project stays about the protocol and the tool contracts rather than becoming another LLM-pipeline demo. Verified end to end against Claude Desktop as a real client.

MCPDOCKERDEVTOOLS