Skip to content
Kitbay
MCP server

Jupyter Server MCP Extension

Runs locallyBSD-3-ClauseUpdated on Oct 12, 202634GitHub starsCan change dataNo API key needed

A Jupyter Server extension from the Jupyter AI Contrib organisation that runs an MCP server inside a running Jupyter Server or JupyterLab.

What it does

A Jupyter Server extension from the Jupyter AI Contrib organisation that runs an MCP server inside a running Jupyter Server or JupyterLab. Which tools it offers is set in Jupyter's configuration: any Python function can be registered by name, and packages can add tools automatically through entry points, for example notebook, file and git tools from jupyter-ai-tools or JupyterLab commands. It listens over streamable HTTP on a local port, and a stdio proxy, runnable with uvx, finds the right running server from the client's working folder. The README gives setup for Claude Code, Codex, OpenCode and Gemini CLI. Jupyter AI uses it as its built-in MCP server.

Capabilities

  • Expose Python functions as MCP tools from a running Jupyter Server
  • Discover tools from installed packages through entry points
  • Bridge stdio clients to the Jupyter MCP endpoint
  • Serve notebook, file and git tools when jupyter-ai-tools is configured