Python Notebook MCP

What it does

This server allows AI assistants to interact with Jupyter Notebook files (.ipynb) on your local machine. It provides a programmatic bridge for LLMs to manage the full lifecycle of a notebook, from creation and cell manipulation to reading outputs.

Tools

• initialize_workspace: Sets the absolute path for the workspace (required first step).
• list_notebooks: Lists all .ipynb files within the workspace.
• create_notebook: Creates a new, empty Jupyter notebook.
• read_notebook: Reads the entire structure and content of a notebook.
• read_cell: Reads the content and metadata of a specific cell by index.
• edit_cell: Modifies the source content of an existing cell.
• add_cell: Adds a new code or markdown cell.
• read_cell_output: Reads the output of a specific code cell.
• read_notebook_outputs: Reads all outputs from all code cells in a notebook.

Installation

To install via Claude Desktop using fastmcp:

fastmcp install server.py --name "Jupyter Notebook MCP"

For manual configuration in claude_desktop_config.json:

{
  "mcpServers": {
    "jupyter": {
      "command": "/path/to/python-notebook-mcp/.venv/bin/python",
      "args": ["/path/to/python-notebook-mcp/server.py"]
    }
  }
}

Supported hosts

• Claude Desktop
• Cursor
• Windsurf