Skip to content
Kitbay
MCP server

Jupyter MCP Server

Local or remoteBSD-3-ClauseUpdated on Nov 11, 20261,323GitHub starsCan change dataNeeds an API key

Datalayer's MCP server for working in Jupyter notebooks in real time.

What it does

Datalayer's MCP server for working in Jupyter notebooks in real time. An agent can browse the Jupyter file system and kernels, open, create and switch between notebooks, read, insert, edit, move and delete cells, clear outputs, and run cells or arbitrary code, getting back text, images and plots. With JupyterLab and its collaboration extension, changes show up live in the open notebook. It works against any Jupyter server or JupyterHub, can run code on other backends such as Colab, Modal or E2B through an optional extension, and is also offered by Datalayer as a hosted endpoint. Locally it runs with uvx or Docker and needs the Jupyter URL and token.

Capabilities

  • Open, create and switch between notebooks
  • Insert, edit, move and delete cells
  • Run cells or code on the notebook kernel and return multimodal output
  • Browse the Jupyter file system and kernels
  • Run code on other sandbox backends through an extension