Notebook Intelligence
Agentic AI coding in JupyterLab with Claude Code, Copilot, Ollama & OpenAI – MCP, Skills, admin policies
Notebook Intelligence is the most capable JupyterLab AI extension we've tested — deep Claude Code integration, multi-provider support, and granular admin controls make it essential for data teams on JupyterHub. The feature depth may overwhelm casual users who just want quick code completion, but for managed deployments it's unmatched.
- Data scientists using JupyterLab wanting AI-assisted coding with agentic capabilities
- Teams needing managed AI policies (provider locking, model endpoints) for JupyterHub
- Developers building custom tools via MCP servers, Skills, and Plugins
- Researchers needing offline local models via Ollama
- Users wanting a standalone AI coding tool outside JupyterLab (NBI requires JupyterLab 4)
- Beginners unfamiliar with JupyterLab extension setup and configuration
- Teams without a JupyterLab 4 environment (NBI only supports JL4)
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
In short
Notebook Intelligence — Agentic AI coding in JupyterLab with Claude Code, Copilot, Ollama & OpenAI – MCP, Skills, admin policies. Best for Data scientists using JupyterLab wanting AI-assisted coding with agentic capabilities, Teams needing managed AI policies (provider locking, model endpoints) for JupyterHub, Developers building custom tools via MCP servers, Skills, and Plugins. Free to use.
What's new in Notebook Intelligence
Checked 11 days agoAcross the latest 8 updates: 8 feature updates.
NBI 5.1.0 released: tool-call status cards, custom spinners, security guardrails
Claude agent tool calls render as persistent status cards with inline diffs. Custom spinner verbs. MCP stdio-command allowlist and filesystem token-password check added.
Coding-agent launchers, Codex support, and security hardening
Launch coding-agent CLIs from JupyterLab launcher. Codex chat models now use correct Copilot endpoint. Accessibility and security passes.
Agent-aware chat sidebar in NBI 5.0
Chat sidebar keeps users informed during long Claude turns and syncs open files with agent disk changes.
Skills, MCP, and Plugins promoted to top-level Settings tabs
Each feature gets force-on/force-off/user-choice admin policy for improved management.
NBI 5.0.0: three new Settings tabs, agent-aware UX, admin policies
Skills, MCP, and Plugins tabs with admin policies. Five coding-agent launcher tiles. Real progress feedback. Official mcp SDK replaces fastmcp.
NBI 4.8.0: smarter workspace picker, atomic config saves, GitHub-import gate
Workspace picker is now .gitignore-aware. Config saves are atomic. New admin policy gates GitHub-based Skill imports.
NBI 4.7.0: cell output actions, agent toolbar, Claude launcher tile
Right-click cell outputs for Explain/Ask/Troubleshoot. Image attachments in chat. Claude Code launcher tile. Streaming inline-chat responses.
NBI 4.6.0: Skills management, restructured docs, Windows fixes
Claude Skills management panel added. Documentation restructured. Windows Claude-mode reliability fixes.
Viability Score
How likely is Notebook Intelligence to still be operational in 12 months? Based on 4 signals — momentum (how recently it shipped), wrapper dependency, revenue model, and web presence.
Last calculated: July 2026
How we score →Key Features
- First-class Claude Code integration with sessions, Skills, Plugins, and MCP
- GitHub Copilot chat and completion models
- Ollama local model support (fully offline)
- OpenAI-compatible and LiteLLM-compatible endpoints
- Inline chat on any code cell (Cmd+I/Ctrl+I)
- Agent mode: creates, edits, runs cells, and fixes errors autonomously
- Chat sidebar with tool-call status cards and inline diffs
- Coding-agent launcher tiles from JupyterLab launcher
- Admin policy controls via environment variables (force-on/off/user-choice)
- Sync organization-wide Skills manifest
- MCP server management for custom tools, databases, APIs
- Cell output actions: Explain/Ask/Troubleshoot
- Image attachments in chat
- Streaming inline-chat responses
- Security passes: MCP stdio-command allowlist and filesystem token-password check
About Notebook Intelligence
Notebook Intelligence (NBI) is an open-source JupyterLab extension that brings agentic AI coding directly into your notebook environment. Built for data scientists, researchers, and developers, it offers a first-class interface for Claude Code — including sessions, Skills, Plugins, and Model Context Protocol (MCP) servers — while also supporting GitHub Copilot, Ollama, and any OpenAI- or LiteLLM-compatible endpoint as drop-in alternatives. The extension works through a chat sidebar and inline chat on any code cell (Cmd+I/Ctrl+I), with agent mode that can create notebooks, edit cells, run them, and fix errors autonomously. Recent updates (NBI 5.0-5.1) introduced tool-call status cards with inline diffs, custom spinners, coding-agent launcher tiles, and a hardened security layer including MCP stdio-command allowlists and filesystem token-password checks. Administrators can enforce provider, model, and endpoint policies via environment variables — force-on, force-off, or user-choice — making it suitable for managed JupyterHub deployments. The extension promotes Skills, MCP Servers, and Plugins to top-level Settings tabs with the same policy triad. NBI is designed to be extended: users can connect MCP servers to expose custom tools, manage Skills and Plugins from the UI, or sync an organization-wide manifest. It also supports cell output actions (Explain/Ask/Troubleshoot), image attachments in chat, and streaming inline-chat responses. The extension is licensed under GPL-3.0 and maintained by Mehmet Bektaş with community contributors. What sets NBI apart is its comprehensive Claude Code agentic support within JupyterLab, combined with flexible provider choices (Copilot, Ollama, OpenAI, LiteLLM, vLLM) and strong admin controls. Unlike standalone AI coding tools, NBI lives inside the notebook environment, keeping your workflow integrated. It's ideal for teams that need managed AI policies without leaving JupyterLab.
Behind the Verdict
Pick Notebook Intelligence when you live in JupyterLab and need an AI coding assistant that actually works inside your notebook — not as a separate window or terminal. It's especially strong for teams running JupyterHub, because the admin policy system (force-on provider, locked models, org-managed Skills) gives you control over what models and tools users can access. The latest 5.x updates make agentic workflows much more transparent: tool-call status cards show you exactly what the agent is doing, and the security guardrails (MCP allowlists, token-password checks) are welcome for enterprise rollouts. Where it bites: NBI only works with JupyterLab 4, so upgrading an older deployment is a prerequisite. Beginners may find the setup — installing pip package, configuring providers, and possibly setting up MCP servers — more involved than a cloud-based AI coding tool. The extension is open-source (GPL-3.0), so there's no licensing cost, but self-support and community help are the norm. The closest alternative is Continue.dev, which also offers multi-provider AI coding in IDEs but lacks the Jupyter-native agentic depth and admin policy surface that NBI provides. In practice, NBI's Claud Code integration is the deepest we've seen in Jupyter — sessions, Skills, Plugins, and MCP all managed from the lab. It's a specialized tool for a specific audience, and for that audience it's excellent. For general-purpose AI coding outside notebooks, stick with GitHub Copilot or Cursor.
Researching Notebook Intelligence? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Use Cases
- Chat with Claude Code inside JupyterLab to debug code and get explanations without leaving the notebook.
- Use inline chat to generate and edit code cells with natural language prompts.
- Run agent mode to automatically create, execute, and fix notebooks from a high-level description.
- Connect local Ollama models for fully offline AI coding assistance.
- Lock provider and model choices across an organization via environment variable policies.
Models Under the Hood
as of 2026-07-16
Limitations
- Notebook Intelligence is a JupyterLab extension that requires a JupyterLab environment.
- It does not include hosted models—users must bring their own API keys or local models.
- Some features like Skills and MCP servers require user configuration.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Integrations
Resources & Guides
Official links
Tools that pair well with Notebook Intelligence
Common stack mates teams adopt alongside Notebook Intelligence, with the specific reason each pairing earns its keep.
Featured Head-to-Head Comparisons
Alternatives to Notebook Intelligence
View all
Cosine Genie
Coding AI trained on production code for maintainable software.
Amazon CodeWhisperer
AI coding assistant with deep AWS integration and agentic automation
Frequently Asked Questions
Categories
Best-of guides
Used Notebook Intelligence? Help shape our editorial sentiment research.