1. Conduid
  2. Documentation
  3. Notebooklm MCP
MCP server · Documentation

Notebooklm MCP

Bridge Google NotebookLM with Claude via MCP. Full programmatic access to create notebooks, manage sources, and generate AI-powered content — podcasts, videos, slides, mind maps, quizzes, flashcards, and reports — all through natural language.

Unclaimed last commit 8 months ago claudeclaude-codeclaude-desktopmcpnotebooklm
44Fair

Scored 4 hours ago · breakdown

About Notebooklm MCP

Notebooklm MCP is an MCP server published by alfredang in the Documentation category: bridge Google NotebookLM with Claude via MCP. Full programmatic access to create notebooks, manage sources, and generate AI-powered content — podcasts, videos, slides, mind maps, quizzes, flashcards, and reports — all through natural language.

The repository has 57 stars and 35 forks, with the last commit 8 months ago. Six months or more without a commit doesn't mean the server is broken, but check the open issues before depending on it in production.

Install

From source
git clone https://github.com/alfredang/notebooklm-mcp

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Ask AI

✦

Ask AI about Notebooklm MCP

Powered by Claude · Grounded in docs

I know everything about Notebooklm MCP. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

Static checks derived from the repository 2 hours ago. Heuristics on the source, not an audit.

  • ✓README presentThe repository ships documentation.
  • !License declaredNo license file. Legally you have no right to use it in a product.
  • !Tests presentNo tests found. Behaviour changes may go unnoticed between versions.
  • ✓Dependencies pinnedA lockfile pins dependency versions.
  • ✓No dynamic code executionSource scan found no eval, dynamic Function, or shell execution.

1 check has not been run yet.

README

NotebookLM MCP for Claude (Desktop & CLI)

A powerful MCP (Model Context Protocol) server that brings Google NotebookLM into Claude Desktop and Claude Code.

Features

  • Research: List and create notebooks
  • Content: Add URLs, text, and files as sources
  • Generation: Create Podcasts, Videos, Slides, Mind Maps, Infographics, Quizzes, Flashcards, and Reports
  • Natural Interaction: Chat directly with your sources using Claude's reasoning

Prerequisites

1. Install uv (Python Package Manager)

# Using curl
curl -LsSf https://astral.sh/uv/install.sh | sh

# Or with Homebrew (macOS)
brew install uv

Default install location: ~/.local/bin/uv

# Using PowerShell
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

# Or with Scoop
scoop install uv

# Or with winget
winget install --id=astral-sh.uv -e

Default install location: %USERPROFILE%\.local\bin\uv.exe

Verify installation:

uv --version

2. Clone and Install Dependencies

# Clone the repository
git clone https://github.com/alfredang/notebooklm-mcp.git

# Navigate to the project folder
cd notebooklm-mcp

# Install dependencies (includes notebooklm-py and fastmcp)
uv sync

This will:

  • Create a .venv virtual environment
  • Install notebooklm-py (Python client for NotebookLM API)
  • Install fastmcp (MCP server framework)

Note: These dependencies are required for both Claude Desktop and Claude Code.


Step 1: Authenticate with NotebookLM

NotebookLM uses browser-based authentication. You must login once to save your session cookies.

cd notebooklm-mcp
uv run notebooklm login

What happens:

  1. A browser window will open automatically
  2. Log in to your Google account
  3. Navigate to NotebookLM if not redirected automatically
  4. Wait until the terminal displays "Success"
  5. Close the browser

Verify authentication:

uv run python -c "
from notebooklm import NotebookLMClient
import asyncio
async def test():
    client = await NotebookLMClient.from_storage()
    async with client:
        notebooks = await client.notebooks.list()
        print(f'Authenticated! Found {len(notebooks)} notebooks.')
asyncio.run(test())
"

You should see: Authenticated! Found X notebooks.


Step 2: Test the MCP Server

Before configuring Claude, verify the server starts correctly:

cd notebooklm-mcp
uv run python server.py

Expected output:

Starting NotebookLM MCP server...
NotebookLM client initialized successfully
Starting MCP server 'NotebookLM' with transport 'stdio'

Press Ctrl+C (or Cmd+C on Mac) to stop the server after confirming it works.


Step 3: Setup for Claude Desktop

3.1 Find Your Paths

You'll need two paths for the configuration:

Find your uv path:

which uv

Example output: /Users/yourname/.local/bin/uv

where uv

Example output: C:\Users\yourname\.local\bin\uv.exe

Find your project path:

cd notebooklm-mcp && pwd

Example output: /Users/yourname/projects/notebooklm-mcp

cd notebooklm-mcp; (Get-Location).Path

Example output: C:\Users\yourname\projects\notebooklm-mcp

3.2 Open the Config File

From Claude Desktop (Recommended):

  1. Open Claude Desktop
  2. Go to Settings (gear icon) → Developer → Edit Config
  3. This opens claude_desktop_config.json in your default editor

Or manually:

~/Library/Application Support/Claude/claude_desktop_config.json

%APPDATA%\Claude\claude_desktop_config.json

3.3 Add the MCP Server Configuration

Important: Replace <UV_PATH> and <PROJECT_PATH> with your actual paths from Step 3.1

{
  "mcpServers": {
    "notebooklm": {
      "command": "<UV_PATH>",
      "args": [
        "--directory",
        "<PROJECT_PATH>",
        "run",
        "python",
        "server.py"
      ]
    }
  }
}

Example with real paths:

{
  "mcpServers": {
    "notebooklm": {
      "command": "/Users/yourname/.local/bin/uv",
      "args": [
        "--directory",
        "/Users/yourname/projects/notebooklm-mcp",
        "run",
        "python",
        "server.py"
      ]
    }
  }
}
{
  "mcpServers": {
    "notebooklm": {
      "command": "<UV_PATH>",
      "args": [
        "--directory",
        "<PROJECT_PATH>",
        "run",
        "python",
        "server.py"
      ]
    }
  }
}

Example with real paths:

{
  "mcpServers": {
    "notebooklm": {
      "command": "C:\\Users\\yourname\\.local\\bin\\uv.exe",
      "args": [
        "--directory",
        "C:\\Users\\yourname\\projects\\notebooklm-mcp",
        "run",
        "python",
        "server.py"
      ]
    }
  }
}

Note: On Windows, use double backslashes (\\) in JSON paths.

3.4 Restart Claude Desktop

Platform How to Restart
macOS Press Cmd+Q to fully quit, then reopen
Windows Right-click tray icon → Quit, then reopen

Look for the hammer icon in the chat input area - this indicates MCP tools are available.

3.5 Verify Connection

In Claude Desktop, type:

List my NotebookLM notebooks

Claude should use the list_notebooks tool and display your notebooks.


Step 4: Setup for Claude Code (CLI)

Prerequisites: Complete Steps 1-2 first (install dependencies with uv sync and authenticate with uv run notebooklm login).

4.1 Add the MCP Server

Replace <PROJECT_PATH> with your actual project path:

claude mcp add notebooklm -- uv --directory <PROJECT_PATH> run python server.py

Example:

claude mcp add notebooklm -- uv --directory /Users/yourname/projects/notebooklm-mcp run python server.py
claude mcp add notebooklm -- uv --directory <PROJECT_PATH> run python server.py

Example:

claude mcp add notebooklm -- uv --directory C:\Users\yourname\projects\notebooklm-mcp run python server.py

4.2 Verify the Server is Added

claude mcp list

You should see notebooklm in the list.

4.3 Test in Claude Code

Start a new Claude Code session:

claude

Then ask:

List my NotebookLM notebooks

Usage Examples

Once configured, use natural language commands in Claude Desktop or Claude Code:

Task Example Command
List notebooks "Show me all my NotebookLM notebooks"
Create notebook "Create a new notebook called 'Research Project'"
Add URL source "Add this URL to my notebook: https://example.com/article"
Generate podcast "Generate a podcast for notebook ID xyz123"
Create slides "Make a slide deck from my 'Research Project' notebook"
Generate mind map "Create a mind map for notebook abc456"
Create quiz "Generate a quiz based on my notebook sources"
Make flashcards "Create study flashcards from this notebook"

Available Tools

Tool Description
list_notebooks List all notebooks in your account
create_notebook Create a new notebook
add_source_url Add a website URL as a source
add_source_text Add raw text as a source
ask_notebook Ask a question based on notebook sources
get_notebook_summary Get summary and key insights
generate_audio_overview Generate a podcast-style audio
generate_video_overview Generate a video overview
generate_slide_deck Generate PowerPoint-style slides
generate_mind_map Generate an interactive mind map
generate_infographic Generate a visual infographic
generate_quiz Generate quiz questions
generate_flashcards Generate study flashcards
generate_summary_report Generate a briefing document
generate_data_table Extract data into a table

Troubleshooting

"Server disconnected" or "Failed to spawn process"

Cause: Claude Desktop can't find uv because it doesn't inherit your shell's PATH.

Solution: Use the full absolute path to uv in the config (see Step 3.1).


"Command not found: uv"

Add to your shell profile (~/.zshrc or ~/.bashrc):

export PATH="$HOME/.local/bin:$PATH"

Then reload:

source ~/.zshrc  # or source ~/.bashrc

Add to your PATH:

  1. Open System Properties → Environment Variables
  2. Under "User variables", edit Path
  3. Add: %USERPROFILE%\.local\bin
  4. Restart your terminal

MCP Server Not Appearing in Claude Desktop

Cause: Invalid JSON in config file or Claude not restarted properly.

Solution:

  1. Validate your JSON at https://jsonlint.com/
  2. Ensure no trailing commas in the JSON
  3. Fully quit and reopen Claude Desktop

"NotebookLM client not initialized"

Cause: Server started before authentication was complete.

Solution:

  1. Run uv run notebooklm login first
  2. Restart Claude Desktop or re-add the MCP server in Claude Code

Check Claude Desktop Logs

# View recent logs
tail -100 ~/Library/Logs/Claude/mcp*.log

# Or open in Finder
open ~/Library/Logs/Claude/
# View logs folder
explorer "$env:APPDATA\Claude\logs"

Remove and Re-add MCP Server (Claude Code)

If issues persist:

claude mcp remove notebooklm
claude mcp add notebooklm -- uv --directory <PROJECT_PATH> run python server.py

Updating

To update the NotebookLM library:

cd notebooklm-mcp
uv sync --upgrade

Project Structure

notebooklm-mcp/
├── server.py          # MCP server implementation
├── pyproject.toml     # Project dependencies
├── README.md          # This file
├── SKILL.md           # Claude Code skill definition
└── .venv/             # Virtual environment (auto-created)

License

MIT License

README fetched from the repository 2 hours ago. The original is authoritative.

Questions

About Notebooklm MCP

How do I install Notebooklm MCP?

No package has been published for this server, so install it from source: clone the repository and follow its README.

Is Notebooklm MCP safe to use with an AI agent?

Its trust score is 44 out of 100 (fair). It passes 3 of the 5 static security checks run so far; the failures are listed above. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Notebooklm MCP still maintained?

The last commit was 8 months ago and 14 open issues. That's long enough that you should check whether the maintainer is responding to issues before depending on it.