Nchan MCP Transport

🚀 Nchan MCP Transport Protocol

A high-performance WebSocket/SSE transport layer gateway for Anthropic's MCP (Model Context Protocol), powered by Nginx, Nchan, and FastAPI. It's suitable for building real-time and scalable AI integrations and communicating with Claude and other LLM agents.

🚀 Quick Start

📦 1. Install the Server SDK

pip install httmcp

🧪 2. Run the Example with Docker

git clone https://github.com/yourusername/nchan-mcp-transport.git
cd nchan-mcp-transport
docker-compose up -d

🛠 3. Define Your Tools

@server.tool()
async def search_docs(query: str) -> str:
    return f"Searching for {query}..."

🧬 4. Expose the OpenAPI Interface (Optional)

If you need to expose the OpenAPI interface, you can do this:

uvicorn server.main:app --reload

Then visit http://localhost:8000/docs in your browser.

✨ Features

The Nchan MCP Transport Protocol provides a real-time API gateway for communication between MCP clients (such as Claude) and their tools and services in the following ways:

• 🧵 WebSocket or Server-Sent Events (SSE)
• ⚡️ Compatible with streamable HTTP
• 🧠 Powered by Nginx + Nchan for low-latency publish/subscribe
• 🛠 Integrates FastAPI for backend logic and the OpenAPI toolchain

✅ Suitable for AI developers to build Claude plugins, LLM agents, or integrate external APIs into the MCP protocol used by Claude.

🧩 Core Features

🧠 Why Use This Project?

MCP allows AI assistants like Claude to communicate with external tools. However:

• Native MCP uses HTTP + SSE, which performs poorly when dealing with long tasks, unstable networks, and high concurrency.
• WebSocket is not natively supported by Claude - this project bridges this gap.
• Pure Python implementations on the server side (such as FastMCP) may not scale under load.

✅ The Nchan MCP Transport Protocol provides:

• Web-level performance (Nginx/Nchan)
• FastAPI-driven backend logic
• Real-time event delivery for Claude clients
• Plug-and-play OpenAPI to Claude integration

🤝 Contribution

Pull requests are welcome! If you have any issues or want to improve the following aspects, please submit an issue:

• Performance optimization
• Deployment improvement
• SDK integration

📄 License

MIT License