notebooklm-skill

You can install notebooklm-skill using several methods depending on your environment requirements.

From a source checkout

Clone the repository and run the provided installer. This uses an isolated environment at ${XDG_DATA_HOME:-~/.local/share}/notebooklm-skill/venv to avoid system Python conflicts. Commands are linked to ${XDG_BIN_HOME:-~/.local/bin}.

Environment Variables:

• NOTEBOOKLM_INSTALL_ROOT: Override the installation directory.
• XDG_BIN_HOME: Override the directory where commands are linked.
• NOTEBOOKLM_PYTHON: Specify a custom Python executable.
• NOTEBOOKLM_INSTALL_EDITABLE=1: Perform an editable install for development.
• NOTEBOOKLM_SKIP_BROWSER=1: Skip Chromium installation (requires manual playwright install chromium later).
• NOTEBOOKLM_SKIP_SKILL=1: Skip installing the Claude Code Skill.
• NOTEBOOKLM_SKIP_AUTH_CHECK=1: Skip authentication checks during install.

From PyPI

Install via pip into a virtual environment and manually install the required Chromium browser.

Using uvx

Run commands in an ephemeral environment without a persistent installation.

# Source checkout
git clone https://github.com/claude-world/notebooklm-skill.git
cd notebooklm-skill
./install.sh

# PyPI
python3 -m venv .venv
source .venv/bin/activate
python -m pip install notebooklm-skill
python -m playwright install chromium

# uvx
uvx --from notebooklm-skill notebooklm-skill --help

Before running workflows, ensure you are authenticated. You can use the profile-aware helper for persistent sessions or a zero-install method for one-off logins.

Persistent Authentication

Use notebooklm-auth to manage sessions. You can specify a specific browser (e.g., Chrome) if you prefer it over the bundled Chromium.

Zero-Install Login

If you do not want a persistent installation, use uvx to run the login command directly.

Profiles

Support for multiple profiles is available via the --profile NAME flag or the NOTEBOOKLM_PROFILE environment variable. Sessions are stored in ~/.notebooklm/profiles/<profile>/storage_state.json (do not access this file directly).

# Standard setup and verification
notebooklm-auth setup
notebooklm-auth verify

# Use local Chrome instead of bundled Chromium
notebooklm-auth setup --browser chrome --fresh

# Zero-install login
uvx --from notebooklm-py notebooklm login

Generate various artifact types from your notebook content.

Supported Types

audio, video, cinematic, slides, report, study-guide, quiz, flashcards, mind-map, infographic, data-table.

Generation Workflow

For long-running media jobs, use --no-wait to get a task ID. You can then list artifacts to find the specific ARTIFACT_ID and download it later.

Output Formats

• Quizzes/Flashcards: Supports json, markdown, or html.
• Slides: Supports pdf or pptx.

Note: Existing files will be rejected unless you use the --force flag.

# Generate slides with specific formatting
notebooklm-skill generate \
  --notebook "AI safety evidence" \
  --type slides --lang zh-TW \
  --slide-format presenter-slides \
  --output ./output/deck.pptx --output-format pptx

# Detached media generation and download
notebooklm-skill generate --notebook NOTEBOOK_ID --type audio --no-wait
notebooklm-skill list-artifacts --notebook NOTEBOOK_ID --type audio
notebooklm-skill download --notebook NOTEBOOK_ID --type audio \
  --artifact-id ARTIFACT_ID --output ./output/podcast.m4a

The notebooklm-skill CLI allows you to create notebooks from mixed sources (URLs, files, and text), inspect them, and ask questions.

Creating a Notebook

You can combine multiple source types in a single command. Use --strict to ensure source integrity.

Adding Sources

To add a single source to an existing notebook, use the specific flags for --url, --file, or --text.

Inspecting and Querying

Use list to see notebooks, list-sources to see what is inside a notebook, and ask to perform grounded queries.

# Create a notebook from mixed sources
notebooklm-skill create \
  --title "AI safety evidence" \
  --sources https://example.com/article https://youtu.be/example \
  --files ./paper.pdf \
  --text-sources "A user-supplied observation" \
  --strict

# Add a single source
notebooklm-skill add-source --notebook "AI safety evidence" --url https://example.com/new

# Query the notebook
notebooklm-skill ask --notebook "AI safety evidence" --query "What findings conflict?"

NotebookLM uses browser-based authentication instead of API keys. The authentication state (session cookies) is stored in a browser state file and should be treated as sensitive data (like a password).

Profile Management

You can use named profiles to separate different Google accounts. Profiles are typically stored in ~/.notebooklm/profiles/<profile>/storage_state.json.

Authentication Tasks

• Setup/Verify: Run setup to start the login flow and verify to check the current session.
• Using Chrome: If the bundled Chromium fails, you can use your system's Google Chrome.
• Clearing Session: Use clear to log out of a specific profile.

Note: Never commit or share the storage_state.json files.

# Basic setup and verification
notebooklm-auth setup
notebooklm-auth verify

# Using system Chrome instead of bundled Chromium
notebooklm-auth setup --browser chrome --fresh

# Managing named profiles (e.g., 'work')
notebooklm-auth --profile work setup
notebooklm-auth --profile work verify
notebooklm-skill --profile work list

# Logging out of a specific profile
notebooklm-auth --profile work clear --yes

# Using an environment variable to select a profile
export NOTEBOOKLM_PROFILE=work