docs/mcp_jupyter_setup.md

Jupyter MCP Server Setup (Workspace)

This project is configured to use datalayer/jupyter-mcp-server inside the local virtual environment.

Installed packages

• jupyter-mcp-server==0.22.1
• jupyterlab==4.4.1
• jupyter-collaboration==4.0.2
• jupyter-mcp-tools>=0.1.4
• ipykernel
• datalayer_pycrdt==0.12.17

Workspace config

• MCP config: .vscode/mcp.json
• Jupyter start script: scripts/start_jupyterlab_for_mcp.ps1
• MCP server start script: scripts/start_jupyter_mcp_server.ps1

Run flow

1. Start JupyterLab:
   • powershell -ExecutionPolicy Bypass -File scripts/start_jupyterlab_for_mcp.ps1
2. Start MCP server (if needed manually):
   • powershell -ExecutionPolicy Bypass -File scripts/start_jupyter_mcp_server.ps1
3. Ensure token in .vscode/mcp.json matches the token used for JupyterLab.

Notes

• Transport is configured as stdio.
• Change URL/token in .vscode/mcp.json for your environment.
• For production or remote use, prefer the upstream Docker/HTTP transport guidance.