jupyter-notebook-mcp

Connects Jupyter Notebook to Claude AI via the Model Context Protocol (MCP), enabling direct AI interaction and notebook control.

Source: GitHub Repository

Category: code-execution-automation-mcp-servers

Tags: jupyter, notebooks, ai-integration, mcp

Features

• Two-way AI Integration: Connects Claude AI to Jupyter Notebook using a WebSocket-based server for real-time communication.
• Cell Manipulation: Insert, execute, edit, and manage notebook cells programmatically.
• Notebook Management: Save notebooks and retrieve notebook and cell information.
• Cell Execution: Run specific cells or all cells in the notebook from Claude or external clients.
• Output Retrieval: Get output content (text or images) from executed cells, with options to limit text length.
• Slideshow Support: Set slide show types for notebook cells to facilitate presentations.
• External Client Testing: Includes a Python-based external client for interactive or batch testing of all available functions.
• Compatibility: Supports only Jupyter Notebook 6.x (not compatible with Jupyter Lab, v7.x, VS Code, or Google Colab).
• Reconnection and Port Management: Automatic reconnection and dynamic port allocation if default port is in use.
• Security: Runs arbitrary code—use with caution and ensure data backup.

Available Tools via Claude AI

• ping: Check server connectivity
• insert_and_execute_cell: Insert and run a cell at a specific position
• save_notebook: Save current notebook
• get_cells_info: Retrieve info about all cells
• get_notebook_info: Retrieve notebook metadata
• run_cell: Execute a specific cell
• run_all_cells: Execute all cells
• get_cell_text_output: Retrieve text output of a cell
• get_image_output: Retrieve image output from a cell
• edit_cell_content: Edit an existing cell's content
• set_slideshow_type: Adjust the slide type for a cell

Limitations

• Works only with Jupyter Notebook 6.x
• Text output from cells is limited to 1500 characters by default
• Does not support advanced Jupyter widget interactions
• Connection may timeout after inactivity

Installation & Usage

• Requires Python 3.12+, uv package manager, and Claude AI desktop app
• Install and configure as described in the repository's README
• Includes instructions for setting up the MCP server, WebSocket server, and kernel integration

License

MIT

Pricing

No pricing information provided. This is an open-source project.