plmbr/notebook-intelligence

What it solves

Notebook Intelligence (NBI) transforms JupyterLab into an AI-powered development environment. It solves the friction of switching between a notebook and an external AI chat by integrating coding assistance, automated notebook manipulation, and multi-model support directly into the JupyterLab interface.

How it works

NBI acts as an extensible framework that connects JupyterLab to various LLM providers. It offers multiple interaction modes:

• Default Mode: Uses configured providers like GitHub Copilot, OpenAI, LiteLLM, or local Ollama models for chat, inline editing, and auto-completion.
• Claude Mode: Integrates the Claude Code CLI and Anthropic API, enabling advanced agentic capabilities, Model Context Protocol (MCP) support, and specialized "skills."
• Agent Mode: A built-in agent capable of interactively creating, editing, and executing notebook cells to fix errors or build workflows.

Who it’s for

• Data scientists and researchers using JupyterLab who want AI-assisted coding.
• Developers who prefer local LLMs (via Ollama) or specific providers like Claude or Copilot.
• Teams requiring administrative control over AI feature policies within their Jupyter environment.

Highlights

• Deep Jupyter Integration: Features include inline chat for code generation, auto-complete in code cells, and the ability to send cell errors or outputs directly to the AI for troubleshooting.
• Claude Code Integration: Full support for Claude Code CLI, including session resumption, permission modes (Default, Accept Edits, Plan, Bypass), and MCP servers.
• Kernel Awareness: Automatically detects the active notebook's language and kernel to ensure generated code is compatible with the specific environment.
• Flexible Provider Support: Compatible with GitHub Copilot, OpenAI, LiteLLM, Ollama, and Anthropic.
• Automatic File Sync: Reloads open tabs when an AI agent modifies files on disk, preventing the need for manual refreshes.