jupyterlab/jupyter-ai — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me an open source JupyterLab extension that lets people use AI agents directly inside their notebooks.

I want a native chat panel where users can talk to agents like Claude, Codex, GitHub Copilot, Gemini, Goose, Kiro, Mistral Vibe, or OpenCode if they have them installed. The agents should be able to understand notebook context, read and write project files, and run terminal commands, but only after asking the user for permission first.

Make it feel practical for data science work. Users should be able to drag notebook cells or files into the chat as context, keep multiple chats going at once, and have the AI help edit notebooks or files without leaving JupyterLab. It should also support real time collaboration for people connected to the same server.

Please keep it flexible so developers can add their own tools, custom MCP servers, and AI personas later. Look up the current JupyterLab and protocol docs online if you need to.

Want more depth? Deep Reverse