Quick Start • Installation • Commands • Troubleshooting • Limitations
NotebookLM
A Claude Code skill for full programmatic access to Google NotebookLM — create notebooks, add sources, chat with citations, and generate podcasts, videos, quizzes and more from the terminal.
🇹🇷 Türkçe için README.tr.md
Quick Start
git clone https://github.com/sametbrr/notebooklm ~/.claude/skills/notebooklm
> "Create a podcast about quantum computing from these three URLs"
That's it — on first use the skill installs the notebooklm-py CLI and walks you through a one-time Google login.
Features
| |
|---|
| Full NotebookLM API | Capabilities beyond the web UI, driven from the terminal |
| All artifact types | Podcast, video, slide deck, infographic, report, quiz, flashcards, mind map, data table |
| Source ingestion | URLs, YouTube, PDFs, audio, video, images |
| Web research | Fast and deep research modes that import sources automatically |
| Chat with citations | Ask questions across sources, save answers as notes |
| Multi-format downloads | mp3, mp4, pdf, pptx, md, csv, json, png |
| One-time login | Browser-based Google auth persisted locally with restrictive permissions |
| Second-brain pattern | Keep a permanent notebook Claude consults before answering |
Requirements
- Claude Code (or any agentskills.io-compatible agent)
- Python 3.10+ (the skill installs 3.12 via Homebrew/apt if needed)
- A Google account with NotebookLM access
- Playwright Chromium (installed automatically during setup)
Installation
git clone https://github.com/sametbrr/notebooklm ~/.claude/skills/notebooklm
Start a new Claude Code session and trigger the skill (e.g. /notebooklm). Setup runs automatically: it creates a venv at ~/.notebooklm-venv, installs notebooklm-py with Playwright, symlinks the CLI to ~/bin/notebooklm, and opens a browser for the one-time Google login. Auth state is stored at ~/.notebooklm/storage_state.json with 600 permissions.
Uninstall
rm -rf ~/.claude/skills/notebooklm ~/.notebooklm-venv ~/.notebooklm
rm -f ~/bin/notebooklm
# remove the PATH line added during setup, if present:
sed -i '' '/HOME\/bin/d' ~/.zshrc
Commands
# Setup & context
notebooklm auth check # Verify authentication
notebooklm list # List notebooks
notebooklm create "Title" # Create a notebook
notebooklm use <notebook_id> # Set active notebook
# Sources & chat
notebooklm source add <url-or-file> # Add a source
notebooklm source list # List sources with status
notebooklm ask "question" # Chat across sources
# Generate & download
notebooklm generate audio "brief" # Podcast (also: video, quiz, report, ...)
notebooklm artifact list # Check generation status
notebooklm download audio ./out.mp3 # Download a finished artifact
notebooklm generate
What it does: starts generation of any artifact type (audio, video, slide-deck, infographic, report, quiz, flashcards, mind-map, data-table) from the active notebook's sources.
Example:
notebooklm generate audio "Focus on the practical implications" --length long
notebooklm artifact wait <artifact_id>
notebooklm download audio ./podcast.mp3
See SKILL.md for the full command reference, generation options, and autonomy rules.
Troubleshooting
command not found: notebooklm — open a new terminal so PATH changes take effect, or run source ~/.zshrc.
Auth/cookie error — the Google session expired. Re-run the login flow in Step 0 of SKILL.md.
Login completes but auth check fails — no SID cookie was captured; delete ~/.notebooklm/browser_profile and ~/.notebooklm/storage_state.json, then log in again and make sure you reach notebooklm.google.com before confirming.
Generation fails or hangs — Google rate limiting. Wait 5–10 minutes and retry with --retry 3.
Limitations
This skill drives an unofficial API — Google can change NotebookLM internals without warning, which may break commands until notebooklm-py updates.
Workaround: pin a known-good notebooklm-py version in the venv and upgrade deliberately.
