Jupyter AI
Open-source JupyterLab extension that brings AI agents such as Claude Code or Codex into notebooks via a chat panel and the %%ai command.
What it does
Jupyter AI adds a chat side panel to JupyterLab where you work with AI agents such as Claude, Codex, GitHub Copilot, Gemini, Goose or Mistral Vibe. Agents connect via the Agent Client Protocol, read and write files, run notebooks and fix errors in cells. You drag files and cells into the chat as context, and several people can chat together. With the optional magic-commands package, %%ai lets you prompt a model directly from a cell. Your own MCP servers, Python functions and JupyterLab commands can be added as tools.
Who it suits
Data science and analytics teams who already work in Jupyter and want AI support inside the notebook instead of switching between editor and chat window.
Data protection for businesses
Jupyter AI is self-hosted software; where data is processed depends on the connected agent or model. That provider’s terms govern usage, training and data processing. Alternatively, you run models via Ollama, vLLM or another OpenAI-compatible server on your own infrastructure, so nothing leaves your network. The project does not offer its own data processing agreement or business plans. Agents must ask for your approval before writing files or running commands.
Limits
Setup and maintenance require Python and Jupyter skills. Nothing works without your own access to an agent or model provider (API key or subscription) or a local model. The project does not mention commercial support.