Master-MCP

One MCP endpoint. All your tools. Fully editable.

Master-MCP is a lightweight proxy MCP server that connects to multiple MCP servers and re-exports their tools through a single MCP endpoint.

It’s designed for agentic workflows where tool catalogs can grow large, messy, and hard to manage.

Why Master-MCP

• Unify MCPs: plug in many MCP servers, expose a single tool list.
• Namespacing: tools are automatically renamed to <server>-<tool>.
• Schema Editor UI: edit tool descriptions/params, hide tools, and curate the catalog.
• Search MCP (optional): regex + BM25 search over tools.
• CodeExec support: generate code-friendly wrappers for Python/TypeScript.

Quick start

pip install -r requirements.txt
python master_mcp_client.py

Then open the schema editor:

open http://127.0.0.1:8001/

Running Master-MCP with an Agent

As a stdio MCP server

Use MCP config with your agent (use path to master_mcp_client.py if not in the root):

{
  "master-mcp": {
        "command": "python",
        "args": ["master_mcp_client.py"]
  }
}

Using HTTP transport through mcp-proxy

npx mcp-proxy \
  --port 3333 \
  python master_mcp_client.py

---

Documentation

• Full documentation & architecture: see FULL_README.md
• Configure external MCPs in mcps.json

Links

• Blogpost: https://towardsdatascience.com/master-mcp-as-a-way-to-keep-mcps-useful-in-agentic-pipelines/
• MCP Python SDK: https://github.com/modelcontextprotocol/python-sdk

---

Roadmap / Ideas

• Code execution mode (with different languages integration)
• Visual UI to enable/disable tools and MCPs at runtime
• Advanced security model (user roles, sandboxing, explicit confirmations)
• Optional HTTP/WebSocket transport in addition to stdio
• Monitoring tools
• Tools heirarchy mechanics
• Tools orchestration
• Add Skills and prompts (just a prompt with no tool execution)

---

🤝 Contribute & Collaborate

If you're interested in contributing, testing new detection methods, or exploring related projects, feel free to contribute and reach me out — collaboration is welcome!

📩 Contact: marvinromson@gmail.com and LinkedIn

---

❤️ Support

If you find this project helpful or use it in your research, you can support further development:
📩 Contact: marvinromson@gmail.com