notebooklm-skill

Give your AI coding assistant a NotebookLM brain

Give your AI coding assistant full programmatic access to Google NotebookLM — with per-project notebooks that automatically link to your codebase and a session wrap-up that saves searchable logs over time.

Works with Claude Code, Cursor, and other AI coding assistants that support instruction files or custom rules.

How it works

This repo has two independent layers:

┌─────────────────────────────────────────────────────┐
│  LAYER 1 — Universal CLI (works everywhere)         │
│  install.sh · notebooklm-py · config.json           │
│  No AI assistant required                           │
└─────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────┐
│  LAYER 2 — AI skills (platform-specific)            │
│  Teach your AI assistant to use the CLI             │
│  skills/ → Claude Code                              │
│  adapters/cursor/ → Cursor                          │
└─────────────────────────────────────────────────────┘

Layer 1 installs a CLI tool (notebooklm) on your machine, sets up per-project config files, and handles authentication. This is the same regardless of which AI assistant you use.

Layer 2 are instruction files for your AI assistant. They teach it how to run the CLI, manage per-project notebooks, detect project context, and save session summaries. Different AI platforms use different file formats — pick the one for your tool.

Compatibility

AI Assistant	Layer 1	Layer 2	Files to use
Claude Code	✅	✅ Native	`skills/notebooklm.md` · `skills/wrapup.md`
Cursor	✅	✅ Adapter	`adapters/cursor/notebooklm.mdc` · `adapters/cursor/wrapup.mdc`
OpenAI Codex	✅	✅ Adapter	`adapters/codex/notebooklm.md` · `adapters/codex/wrapup.md`
GitHub Copilot	✅	⚠️ Partial	Use Layer 1 + manual instructions
Any other agent	✅	⚠️ Manual	See adapters/README.md

Missing your tool? The adapters guide explains how to port the skills to any platform. PRs welcome.

Prerequisites

Python 3.10+
A Google account with access to NotebookLM
uv (strongly recommended) — a fast Python package manager that avoids common install issues on macOS and Linux. Install it with:
```
curl -LsSf https://astral.sh/uv/install.sh | sh   # macOS / Linux
pip install uv                                      # any OS
```
If you skip this, the installer falls back to pip automatically.

Step 1 — Install the CLI (everyone)

git clone https://github.com/ibaifernandez/notebooklm-skill.git
cd notebooklm-skill
chmod +x install.sh && ./install.sh

The script:

Detects Python 3.10+ (prefers 3.12 for best wheel compatibility)
Installs uv if missing
Creates ~/.notebooklm-venv with notebooklm-py
Installs Playwright Chromium (~200 MB, cached after first run)
Adds a notebooklm symlink to ~/bin

Verify:

notebooklm --version

Step 2 — Install the skills for your AI assistant

Claude Code

Claude Code skills must be installed as directories containing a SKILL.md file:

mkdir -p ~/.claude/skills/notebooklm
mkdir -p ~/.claude/skills/wrapup
cp skills/notebooklm.md ~/.claude/skills/notebooklm/SKILL.md
cp skills/wrapup.md ~/.claude/skills/wrapup/SKILL.md

Skill directory: ~/.claude/skills/ on macOS/Linux, %USERPROFILE%\.claude\skills\ on Windows. Create it if it doesn't exist: mkdir -p ~/.claude/skills Each skill lives in its own subdirectory with a SKILL.md file inside — a flat .md file at the root will not be picked up by Claude Code.

Cursor

mkdir -p .cursor/rules
cp adapters/cursor/notebooklm.mdc .cursor/rules/
cp adapters/cursor/wrapup.mdc .cursor/rules/

Run this inside each project where you want NotebookLM integration, or copy to a global rules location if your Cursor version supports it.

OpenAI Codex

Codex supports the same /skill-name trigger as Claude Code. Install the skills in your Codex skills directory (Personal scope) and invoke them with /notebooklm and /wrapup in chat.

Skills directory: check Codex → Settings to find where Personal skills are stored on your system.

cp adapters/codex/notebooklm.md ~/.codex/skills/
cp adapters/codex/wrapup.md ~/.codex/skills/

Then in Codex chat: /notebooklm init, /notebooklm sync, /wrapup.

Other assistants

See adapters/README.md.

Step 3 — Authenticate (first time only)

Authentication is handled by your AI assistant during the first /notebooklm (Claude Code) or @notebooklm (Cursor) invocation. You can also trigger it manually:

Claude Code: type /notebooklm in any project — Claude detects missing auth and runs the login flow.

Cursor: open Cursor in any project and type @notebooklm authenticate in chat.

Manually (any tool):

# The built-in "notebooklm login" requires interactive input that most AI
# bash tools don't support. Use this script instead:

cat > /tmp/nlm_login.py << 'PYEOF'
import json, time
from pathlib import Path
from playwright.sync_api import sync_playwright

STORAGE_PATH = Path.home() / ".notebooklm" / "storage_state.json"
PROFILE_PATH = Path.home() / ".notebooklm" / "browser_profile"
SIGNAL_FILE  = Path("/tmp/nlm_save_signal")

SIGNAL_FILE.unlink(missing_ok=True)
STORAGE_PATH.parent.mkdir(parents=True, exist_ok=True)

with sync_playwright() as p:
    browser = p.chromium.launch_persistent_context(
        user_data_dir=str(PROFILE_PATH),
        headless=False,
        args=["--disable-blink-features=AutomationControlled"],
    )
    page = browser.pages[0] if browser.pages else browser.new_page()
    page.goto("https://notebooklm.google.com/")
    print("Sign in to Google, then run: touch /tmp/nlm_save_signal")
    while not SIGNAL_FILE.exists():
        time.sleep(1)
    storage = browser.storage_state()
    STORAGE_PATH.write_text(json.dumps(storage))
    print(f"Saved {len(storage.get('cookies',[]))} cookies to {STORAGE_PATH}")
    browser.close()
SIGNAL_FILE.unlink(missing_ok=True)
PYEOF

source ~/.notebooklm-venv/bin/activate
python3 /tmp/nlm_login.py &

# Once you're signed in and on notebooklm.google.com:
touch /tmp/nlm_save_signal

# Verify
notebooklm auth check
notebooklm list

Cookies expire after a few days. Re-run the login flow whenever notebooklm auth check fails.

Usage

Link a project to a notebook

Claude Code: /notebooklm init

Cursor: @notebooklm init this project

OpenAI Codex: /notebooklm init

Manual:

cd your-project
notebooklm create "$(basename $(pwd))" --json
# Note the notebook ID, then:
mkdir -p .notebooklm
cat > .notebooklm/config.json << EOF
{
  "notebook_id": "<ID FROM ABOVE>",
  "notebook_title": "$(basename $(pwd))",
  "project_path": "$(pwd)",
  "created_at": "$(date +%F)",
  "auto_sources": ["README.md", "docs/"],
  "live_url": null
}
EOF
notebooklm use <ID>
notebooklm source add README.md

Your AI assistant (Claude Code, Cursor, or Codex) handles all of this automatically when you use the init command. The manual steps are shown here so you understand what's happening and can use the CLI directly if you prefer.

Ask questions about your project

# In Claude Code or Codex:
/notebooklm
> What are the main architectural decisions in this codebase?

# In Cursor chat:
@notebooklm What are the main architectural decisions?

# Direct CLI:
notebooklm ask "What are the main architectural decisions?"

Generate artifacts

notebooklm generate audio "Focus on the architecture decisions"
notebooklm generate report --format briefing-doc
notebooklm generate quiz --difficulty medium
notebooklm artifact wait <artifact_id>
notebooklm download audio ./podcast.mp3

Artifact types: audio (podcast), video, report, quiz, flashcards, slide-deck, infographic, mind-map.

Sync sources after changes

# Claude Code / Codex:   /notebooklm sync
# Cursor:                @notebooklm sync sources
# Direct CLI:
notebooklm source add ./docs/new-doc.md

End-of-session wrap-up

# Claude Code / Codex:   /wrapup
# Cursor:                @wrapup

Saves memories, writes a session log, and pushes it to the project's notebook.

Project config

Each project stores its notebook at .notebooklm/config.json:

{
  "notebook_id": "9d812075-8f79-4a8b-b398-fdb1ebf91eb2",
  "notebook_title": "my-project",
  "project_path": "/Users/me/projects/my-project",
  "created_at": "2026-04-26",
  "auto_sources": ["README.md", "CLAUDE.md", "docs/"],
  "live_url": "https://my-project.com"
}

A global registry at ~/.notebooklm/registry.json maps project paths to notebook IDs, so the right notebook loads automatically even without the config file.

File layout

~/.notebooklm/
  storage_state.json    ← auth cookies (never commit this)
  browser_profile/      ← Playwright browser profile
  registry.json         ← global map: project path → notebook ID

<your-project>/
  .notebooklm/
    config.json         ← per-project notebook config
    sessions/           ← session logs pushed to NotebookLM

Supported OS

Platform	Status	Notes
macOS (Apple Silicon)	✅ Tested	Use `uv` — system pip has `ensurepip` issues on some Homebrew setups
macOS (Intel)	✅ Should work	`uv` recommended
Linux (Ubuntu / Debian)	✅ Should work	`sudo apt install python3.12` if needed
Windows (WSL2)	🟡 Likely works	Not yet tested; use WSL2 + `uv`
Windows (native)	🟡 Experimental	Paths will differ; use full venv path

CLI reference

notebooklm auth check           Check authentication status
notebooklm list                 List your notebooks
notebooklm create "Title"       Create a notebook
notebooklm use <id>             Set active notebook
notebooklm status               Show active notebook
notebooklm source add <file>    Add a file source
notebooklm source add <url>     Add a URL source
notebooklm source list          List sources
notebooklm ask "question"       Chat with your notebook
notebooklm generate audio       Generate podcast
notebooklm generate report      Generate report
notebooklm generate quiz        Generate quiz
notebooklm artifact list        List generated artifacts
notebooklm download audio <f>   Download audio artifact

Full reference: notebooklm --help

Built on

This skill is built on top of notebooklm-py by @teng-lin — an unofficial Python CLI for the NotebookLM web interface.

The idea of wrapping notebooklm-py in an AI coding assistant skill was inspired by community experimentation around the library. The per-project notebook architecture, cross-platform installer, and multi-agent adapter system are original additions.

Important: NotebookLM has no official public API. notebooklm-py drives the web UI via Playwright. Google may change things without warning. If something breaks, update the library:

uv pip install --upgrade notebooklm-py --python ~/.notebooklm-venv/bin/python

Contributing

PRs welcome — especially adapters for new AI platforms. See adapters/README.md for porting instructions.

To report a bug: OS, Python version, notebooklm --version, the failing command, and the error output.

MIT License · Built by Ibai Fernández · AGLAYA

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
adapters		adapters
images		images
skills		skills
.gitignore		.gitignore
LICENSE		LICENSE
LINKEDIN.md		LINKEDIN.md
PRODUCT-HUNT.md		PRODUCT-HUNT.md
README.md		README.md
install.sh		install.sh

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

notebooklm-skill

How it works

Compatibility

Prerequisites

Step 1 — Install the CLI (everyone)

Step 2 — Install the skills for your AI assistant

Claude Code

Cursor

OpenAI Codex

Other assistants

Step 3 — Authenticate (first time only)

Usage

Link a project to a notebook

Ask questions about your project

Generate artifacts

Sync sources after changes

End-of-session wrap-up

Project config

File layout

Supported OS

CLI reference

Built on

Contributing

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Folders and files

Latest commit

History

Repository files navigation

notebooklm-skill

How it works

Compatibility

Prerequisites

Step 1 — Install the CLI (everyone)

Step 2 — Install the skills for your AI assistant

Claude Code

Cursor

OpenAI Codex

Other assistants

Step 3 — Authenticate (first time only)

Usage

Link a project to a notebook

Ask questions about your project

Generate artifacts

Sync sources after changes

End-of-session wrap-up

Project config

File layout

Supported OS

CLI reference

Built on

Contributing

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages