What it does

Jupyter AI brings agent-based AI into JupyterLab through a native chat interface. It works with agents such as Claude, Codex, GitHub Copilot, Gemini, Goose, Kiro, Mistral Vibe and OpenCode over the Agent Client Protocol, and it detects agents automatically once their dependencies are installed. Agents can read and write files, run terminal commands and work with notebooks through a built-in Jupyter MCP server. A permission system asks for your approval before an agent writes files or executes commands. You can keep several chats going at once, drag files or notebook cells in as context, and collaborate in real time with other people connected to the same server. It is meant to be extended. You can add custom MCP servers for domain-specific tools, and developers can register their own AI personas through the entry points API. The project says it is under incubation as part of the JupyterLab organization and is released under the BSD-3-Clause license.

Capabilities

Chat interface inside JupyterLab
Works with several agents through the Agent Client Protocol
Built-in Jupyter MCP server for notebook access
Approval prompts before file writes or commands
Multiple chats with files or notebook cells as context
Custom MCP servers and AI personas

When to use it

You work in JupyterLab notebooks and want an AI agent to read and edit them
You want to pick between several coding agents from one interface
You want approval guardrails before an agent changes files or runs commands
You want to extend the assistant with your own MCP servers or personas