notebooklm-wrapper 0.1.4

Pythonic wrapper for NotebookLM via notebooklm-mcp-cli MCP server

NotebookLM Wrapper

Pythonic wrapper for Google NotebookLM via the MCP (Model Context Protocol). Connects to the notebooklm-mcp server and provides a clean, typed interface to all NotebookLM functionality.

Features

• Clean Python API - No subprocess or CLI calls
• Full type safety - Pydantic models for all data
• Sync + Async - Use either programming style
• 28 operations - Notebooks, sources, chat, research, studio, sharing, and more
• Auth handled by server - Run nlm login once, then use the wrapper

Requirements

• Python 3.11+
• notebooklm-mcp-cli (installed automatically)

Installation

From PyPI (once published):

pip install notebooklm-wrapper

From source (development or unreleased):

pip install git+https://github.com/ai-chitect/notebooklm-wrapper.git
# or, from a local clone:
pip install -e .

Publishing to PyPI: Run ./scripts/publish.sh (runs tests, builds, then publishes). Use ./scripts/publish.sh --test for TestPyPI. Requires hatch and PyPI credentials.

Prerequisites: Authenticate with NotebookLM (one-time setup):

pip install notebooklm-mcp-cli
nlm login  # Opens browser for authentication

Quick Start

from notebooklm_wrapper import NotebookLMClient

# Initialize client (uses default profile from nlm login)
client = NotebookLMClient()

# List notebooks
notebooks = client.notebook.list()
for nb in notebooks:
    print(f"{nb.title} ({nb.id})")

# Create notebook
notebook = client.notebook.create(title="My Research")

# Add sources
client.source.add(
    notebook.id,
    "url",
    url="https://example.com/article",
)

# Ask questions
response = client.chat.ask(notebook.id, "What are the main points?")
print(response.answer)
for citation in response.citations:
    print(f"  - {citation.source_title}")

Async Usage

import asyncio
from notebooklm_wrapper import AsyncNotebookLMClient

async def main():
    client = AsyncNotebookLMClient()
    notebooks = await client.notebook.list()
    print(f"Found {len(notebooks)} notebooks")

    # Create and query
    notebook = await client.notebook.create(title="Async Research")
    response = await client.chat.ask(notebook.id, "Summarize the key findings")
    print(response.answer)

asyncio.run(main())

Multi-Profile

# Use specific profile (set via NOTEBOOKLM_MCP_PROFILE when spawning server)
client = NotebookLMClient(profile="work")
notebooks = client.notebook.list()

# Async with profile
async_client = AsyncNotebookLMClient(profile="personal")

Multi-User / Web App

For multi-tenant apps (e.g. on Google Cloud) where each user has their own NotebookLM credentials stored in your DB (encrypted) and obtained via OAuth, use config_dir to isolate credentials per user. The wrapper spawns one MCP server process per client; with a unique config_dir per user, that process uses a dedicated credential store under config_dir/.notebooklm-mcp-cli.

1. Store credentials: After OAuth (or user-provided cookies), encrypt and store the cookie string in your DB keyed by user id.
2. Per request/session: Create a user-specific directory (or use a persistent path per user), decrypt credentials from your DB, then create a client and inject tokens:

import os
from notebooklm_wrapper import AsyncNotebookLMClient

async def notebooklm_client_for_user(user_id: str, decrypted_cookies: str):
    # Use a dedicated dir per user so the MCP server stores credentials there
    config_dir = f"/app/data/users/{user_id}"
    os.makedirs(config_dir, exist_ok=True)

    client = AsyncNotebookLMClient(
        profile=user_id,
        config_dir=config_dir,
    )
    # Inject credentials so the server can use them (first time or refresh)
    await client.auth.save_tokens(cookies=decrypted_cookies)
    return client

# Then use the client as usual
async def handle_request(user_id: str, ...):
    cookies = get_encrypted_cookies_from_db(user_id)
    decrypted = decrypt(cookies)
    client = await notebooklm_client_for_user(user_id, decrypted)
    try:
        notebooks = await client.notebook.list()
        # ...
    finally:
        await client.disconnect()

See docs/multi-user-credentials-design.md for how credential isolation works and design details.

API Reference

Resource	Methods
client.notebook	list(), get(id), describe(id), create(title), rename(id, title), delete(id, confirm=True)
client.source	add(notebook_id, type, url=...), list_drive(notebook_id), sync_drive(ids, confirm=True), delete(id, confirm=True), describe(id), get_content(id)
client.chat	ask(notebook_id, query), configure(notebook_id, goal=..., response_length=...)
client.research	start(query, source="web", mode="fast"), status(notebook_id), import_sources(notebook_id, task_id)
client.studio	create(notebook_id, type, confirm=True), status(notebook_id), delete(notebook_id, artifact_id, confirm=True)
client.share	status(notebook_id), set_public(notebook_id, is_public), invite(notebook_id, email, role="viewer")
client.download	artifact(notebook_id, type, output_path)
client.note	create(notebook_id, content, title=...), list(notebook_id), update(notebook_id, note_id, ...), delete(notebook_id, note_id, confirm=True)
client.auth	refresh(), save_tokens(cookies, ...)
client.export	to_docs(notebook_id, artifact_id), to_sheets(notebook_id, artifact_id)

Error Handling

All operations raise NotebookLMError subclasses on failure:

from notebooklm_wrapper import (
    NotebookLMClient,
    AuthenticationError,
    NotFoundError,
    ValidationError,
    RateLimitError,
    GenerationError,
)

client = NotebookLMClient()

try:
    notebooks = client.notebook.list()
except AuthenticationError:
    print("Run 'nlm login' to authenticate")
except NotFoundError as e:
    print(f"Not found: {e}")
except RateLimitError as e:
    print(f"Rate limited. Retry after: {e.retry_after}s")

Error messages include the tool name for easier debugging (e.g. [notebook_list] Please login first).

Development

# Clone the repository
git clone https://github.com/ai-chitect/notebooklm-wrapper.git
cd notebooklm-wrapper

# Create virtual environment
python -m venv .venv
source .venv/bin/activate   # Linux/macOS
# .venv\Scripts\activate    # Windows

# Install in editable mode with dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Run tests with coverage
pytest --cov=notebooklm_wrapper --cov-report=term-missing

# Lint
ruff check src/ tests/

# Format
black src/ tests/
isort src/ tests/

# Type check
mypy src/

# Publish to PyPI (after tests pass)
./scripts/publish.sh

# Test OAuth + list notebooks (integration-style; opens browser for login)
python scripts/test_oauth_list_notebooks.py
# Reuse credentials in a directory (skip login next time):
python scripts/test_oauth_list_notebooks.py --config-dir ./tmp_oauth_test
python scripts/test_oauth_list_notebooks.py --config-dir ./tmp_oauth_test --skip-login

# Test paste-cookie flow (use a file for long cookie strings):
echo "SID=...; __Secure-1PSID=...; ..." > cookies.txt   # paste from browser
python scripts/test_cookie_list_notebooks.py --cookie-file cookies.txt
# Or prompt for short cookie, or stdin:
python scripts/test_cookie_list_notebooks.py
cat cookies.txt | python scripts/test_cookie_list_notebooks.py --stdin

Acknowledgments

This wrapper connects to the notebooklm-mcp-cli MCP server by Jacob Ben-David, which provides the underlying NotebookLM integration. That project is MIT-licensed and compatible with this wrapper.

License

MIT