NotebookLM
Unofficial Python library and CLI for Google Gemini Notebook (formerly NotebookLM), including AI-agent skill steng-lin/notebooklm-py
- Stars
- 20.1k
- Forks
- 2.7k
- Licence
- MIT
- Language
- Python
- Last update
- 4 days ago
- Type
- Library
What it is
notebooklm-py is an unofficial Python package that lets you control Google's Gemini Notebook (formerly NotebookLM) from code or the command line, including some features the web interface does not expose. It also ships a skill that lets AI agents such as Claude Code and Codex use notebooks to answer questions from your sources. People use it to automate source imports, generate and download artifacts like audio overviews and quizzes, and get cited answers from their own documents.
- Bulk-import URLs, PDFs, YouTube videos, or Drive files into a notebook and ask questions with cited answers
- Generate audio overviews, slide decks, quizzes, or flashcards from a set of sources and download them locally
- Export mind maps as JSON, flashcards and quizzes in several formats, or data tables as CSV
- Let an AI agent such as Claude Code query a notebook of project docs before answering
Great for
- Developers who want to script Gemini Notebook workflows in Python
- People building AI agents that need answers grounded in their own documents
- Researchers bulk-importing sources and exporting generated artifacts
- Prototypes, research, and personal projects, as the README recommends
Not ideal for
- Production systems that need guaranteed uptime, because it relies on undocumented Google APIs that can change without notice
- Anyone needing official Google support or a stable contract, since it is a community project not affiliated with Google
- Heavy-volume workloads, because the README warns that rate limits may throttle heavy usage
How to get it
- 1
Install with a package manager
You'll need: Python 3.10 or newer (the README badge lists 3.10 through 3.15)
Install from PyPI
pip install notebooklm-py - 2
Run it
- Install the package with pip
- Authenticate with your Google account as described in the README or docs
- Create a notebook and add a source using the CLI or Python API
- Ask a question against the notebook to get a cited answer
- 3
Get the most out of it
- Install the bundled skill with the notebooklm skill install command for Claude Code or .agents skill directories
- Use ask --json to get structured, source-grounded answers you can parse in scripts
- Use the MCP server documented in docs/mcp-guide.md to expose a notebook to other agents
- Run the CLI from an Obsidian vault root to save downloaded artifacts as local files
Downloads
Strengths
- Exposes programmatic access to features the web UI does not offer, such as batch downloads and mind map JSON extraction
- Works through Python, a CLI, and AI agent skills, so it fits several workflows
- Open source under the MIT license with a visible test badge and an active repository
Trade-offs
- Built on undocumented Google APIs, so it may break without warning
- Subject to rate limits, and the README suggests it is best for prototypes and personal use rather than production
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| notebooklm-py (this) | 20.1k | Library | intermediate | 4 days ago | Developers who want to script Gemini Notebook workflows in Python |
| awesome-agent-skills | 36.8k | Curated list | beginner | 4 days ago | Developers who use Claude Code, Codex, Cursor, or Gemini CLI and want ready-made skills |
| baoyu-skills | 27.2k | App | intermediate | 4 days ago | Creators who publish articles to WeChat Official Accounts and want cover and illustration help |
| Claude-Red | 10.2k | Curated list | beginner | 24 days ago | Security professionals who already use Claude and want topic-specific methodology for authorized testing |
| semantic-kernel | 28.7k | Library | intermediate | 4 days ago | Developers already using Azure OpenAI or OpenAI who want a structured SDK for agents |
| agent-framework | 14.5k | Library | intermediate | 4 days ago | Developers building agents or multi-agent workflows they plan to run in production |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.