jupyterlab/jupyter-ai

What it solves

Jupyter AI integrates agentic AI directly into the JupyterLab environment, allowing users to collaborate with AI agents that can interact with their computational notebooks, files, and terminal commands without leaving their workspace.

How it works

It provides a native chat interface within JupyterLab. It uses the Agent Client Protocol (ACP) to connect to various frontier AI agents (such as Claude, Gemini, and Mistral Vibe). These agents can read and write files, execute terminal commands, and interact with notebooks via a built-in Jupyter MCP server. To ensure security, a permission system requires agents to request approval before performing sensitive actions like writing files or executing commands.

Who it’s for

Data scientists, researchers, and developers who use JupyterLab and want to integrate AI agents into their computational workflows to automate tasks and and manage files and code.

Highlights

• Agentic Capabilities: Agents can run terminal commands, read/write files, and interact with notebooks.
• Open Standards: Built on ACP and MCP to avoid vendor lock-in and support a wide ecosystem of compatible agents.
• Contextual Awareness: Supports dragging and dropping files or notebook cells as context for the chat.
• Extensibility: Allows developers to register custom AI personas and add custom MCP servers for domain-specific tools.
• Collaborative: Supports real-time collaboration with other users on the same server.