A Curated List of Open Source MCP Projects and Tools

A Curated List of Open Source MCP Projects and Tools

The Model Context Protocol (MCP) gives AI agents a standard way to reach tools, codebases, and applications. The projects below show where that standard has taken hold. Treat this as a map, not a ranking: each entry names what the project does and where it fits, so you can assemble a workflow from proven components instead of building every capability yourself.

They fall into three groups — framework integrations, developer-experience tools, and automation and testing infrastructure.

Framework and platform integrations

These projects wire MCP into existing frameworks so agents can act on real applications and development workflows.

  • fastapi_mcp — Exposes FastAPI endpoints as MCP tools with minimal setup, carrying authentication and configuration through from the existing app.
  • nuxt-mcp — Adds Nuxt developer tools for route inspection and server-side-rendering debugging, giving models a clearer view of Vite/Nuxt applications.
  • unity-mcp — Bridges to the Unity game engine’s APIs for AI-assisted development: managing assets, controlling scenes, editing scripts, and automating tasks in the editor.

Developer experience and AI-assisted coding

These focus on productivity — turning an LLM into a capable IDE collaborator.

  • context7 — Pulls version-specific documentation and code examples from a codebase and injects them into the model’s context, cutting down on outdated or hallucinated API usage.
  • serena — A toolkit for agent-driven coding with semantic code retrieval and editing, working from program structure rather than plain text matching.
  • Peekaboo — A Swift tool that translates on-screen GUI elements into actionable context, enabling GUI automation for AI assistants.
  • coderunner — Turns an LLM into a local execution partner: it writes and runs code in a preconfigured sandbox, auto-installs dependencies, reads files, and returns outputs or generated artifacts.
  • MCP CLI — A lightweight command-line interface for MCP servers, built for coding agents. It addresses context-window bloat by letting agents discover tools just in time rather than loading every schema upfront. See MCP CLI: Dynamic Tool Discovery for AI Agents for the full walkthrough.

Automation, testing, and orchestration

These provide the infrastructure for running and validating MCP-based automation.

  • n8n-mcp — Extends the n8n automation platform so models can help create, orchestrate, and reason about n8n nodes and workflows.
  • inspector — A tool for testing and debugging MCP servers. It exposes the protocol handshake, tools, resources, prompts, and OAuth flows, and includes a built-in model playground and evaluation runs for catching security or performance regressions before they ship.

Where to start

If you are exposing an existing service, begin with a framework integration (fastapi_mcp, nuxt-mcp). If you are building an agent that writes or reasons about code, pair a discovery layer (MCP CLI) with a semantic toolkit (serena, context7). And whatever you build, keep inspector in the loop — validating a server against the protocol before wiring it into an agent saves far more time than it costs.

This entry was posted in . Bookmark the permalink.
Adam Bernard