jupyter

This MCP server provides a configurable bridge that exposes Jupyter Server functionality as MCP tools. The server runs as a Python-based MCP extension for Jupyter Server and can dynamically load and register Python callables as tools accessible to MCP clients. Tools can be registered manually in configuration using the module:function format, or automatically discovered via entrypoints in registered Python packages. Once started, clients such as Claude or Gemini can connect to the HTTP MCP endpoint to discover available tools and invoke them through the MCP protocol. Typical use cases include performing notebook-related actions, file operations, or other Jupyter-integrated tasks from AI assistants.

Prerequisites:

• Python 3.8+ installed on the host
• pip (Python package manager) available in PATH

Install the MCP extension package for Jupyter Server:

pip install jupyter-server-mcp

Optional: create a configuration for Jupyter Server that enables the MCP extension. For example, ensure your jupyter_config.py includes MCP-related settings as shown in the README's quick start. You can start Jupyter Server with the extension enabled and configured, for example:

jupyter lab --config=jupyter_config.py

If you are developing locally, install in editable mode and run tests as needed:

pip install -e .
pytest tests/ -v

Tips and common considerations:

• Tools are registered via the mcp_tools configuration attribute. Use strings in the format module:function, for example: "os:getcwd" or complex tools from Jupyter-related packages.
• You can enable automatic discovery of tools via entrypoints by setting use_tool_discovery to True (default).
• Ensure the MCP server port is available and not blocked by a firewall. Configure your clients (Claude, Gemini) to point to http://<host>:<port>/mcp as described in the readme.
• If tools fail to register, verify that the target modules are importable in the Python environment running the MCP server and that any external dependencies are installed.
• The extension is designed to integrate with Jupyter Server. For production, consider running behind a reverse proxy and enabling TLS/credentials as appropriate.