MCP Server

The MCP Server empowers AI assistants like Claude to utilize external tools safely, such as running Python code or accessing websites. It enhances data processing for AI understanding and offers detailed error messages, enabling bots to resolve issues independently.

1. Tools offered
2. How to run it
   1. Docker (recommended)
   2. Locally
3. How to connect
4. Learn more about MCP
5. License

Tools offered

The server provides:

• Python: Executes Python code in a secure sandbox, including libraries like NumPy and Pandas.
• Web: Accesses websites, converts pages to Markdown, retrieves raw content, and extracts links.

How to run it

Docker (recommended)

Use docker-compose.yml to configure the server with environment variables like SSE_HOST, SSE_PORT, and USER_AGENT. Start the server with docker compose up. Ideal for LibreChat, configure librechat.yaml with the server's URL.

Running locally

Install uv, create a virtual environment, install dependencies using uv sync, and run the server with arguments like --sse-host, --sse-port, and --user-agent.

How to connect

Choose between network (SSE) for service-based connections or direct (stdio) for local testing.

Learn more about MCP

Explore the MCP Specification, Python SDK, and Example Servers.

License

Licensed under GPLv3. See LICENSE.