Notebooklm Mcp

notebooklm-mcp

An MCP (Model Context Protocol) server that exposes Google NotebookLM as a tool surface for Claude Desktop, Claude Code, and any other MCP-compatible client.

Built on top of notebooklm-py, an unofficial Python API for NotebookLM.

> Warning — Unofficial / Undocumented APIs > > notebooklm-py drives Google's internal NotebookLM endpoints, which are > undocumented and subject to change without notice. Google is not > affiliated with this project. API breakage can happen at any time. > > Recommended: use a throwaway Google account for this integration > rather than your primary account. Rate limits and terms-of-service > restrictions apply.

---

Available Tools

| Tool | Description | |------|-------------| | notebooklm_list_notebooks | List all notebooks in the account | | notebooklm_create_notebook | Create a new notebook | | notebooklm_add_source | Add a URL, YouTube link, or pasted text as a source | | notebooklm_ask | Ask a question against a notebook's sources | | notebooklm_generate_audio_overview | Generate an Audio Overview (podcast deep dive) | | notebooklm_list_artifacts | List all generated artifacts in a notebook | | notebooklm_download_artifact | Download an artifact to a local file |

---

Installation & Setup

1. Install the package

Option A — uvx (recommended, no persistent install needed):

# No install step required; uvx downloads and runs on demand.
# Just make sure uv is installed: https://docs.astral.sh/uv/

Option B — pip into a dedicated environment:

pip install "notebooklm-mcp"

Option C — from source (this repo):

git clone https://github.com/deji-adesoga/notebooklm-mcp
cd notebooklm-mcp
python -m venv .venv
.venv\Scripts\activate        # Windows
# source .venv/bin/activate   # macOS/Linux
pip install -e .

2. Install the Chromium browser (required for login)

playwright install chromium

If you see TypeError: onExit is not a function on Linux, see the notebooklm-py troubleshooting guide.

3. Authenticate (one-time)

notebooklm login

This opens a Chromium browser window. Log in with your (throwaway) Google account. Credentials are saved to ~/.notebooklm/profiles/default/storage_state.json. You only need to do this once — the MCP server reads the stored cookies automatically.

To verify authentication worked:

notebooklm auth check --test

---

Claude Desktop Configuration (Windows MSIX)

The MSIX edition of Claude Desktop reads its MCP config from:

%LOCALAPPDATA%\Packages\Claude_pzs8sxrjxfjjc\LocalCache\Roaming\Claude\claude_desktop_config.json

Open that file (create it if it doesn't exist) and add the notebooklm entry to the mcpServers object.

Option A — uvx (no persistent install)

{
  "mcpServers": {
    "notebooklm": {
      "command": "uvx",
      "args": ["notebooklm-mcp"]
    }
  }
}

Option B — local venv (from-source install)

{
  "mcpServers": {
    "notebooklm": {
      "command": "C:\\Users\\Diggity\\Documents\\Local Repositories\\notebooklm-mcp\\.venv\\Scripts\\python.exe",
      "args": ["-m", "notebooklm_mcp.server"]
    }
  }
}

Restart Claude Desktop after editing the config.

---

Quickstart (once configured)

Once the server is running, Claude can use natural language to drive NotebookLM:

List my notebooks.

Create a notebook called "AI Safety Reading List".

Add https://arxiv.org/abs/2301.13379 as a source and wait for it to index.

Ask "What are the key findings?" in that notebook.

Generate an audio overview for the notebook.

Download the audio to C:/Users/Diggity/Downloads/podcast.mp3.

---

Artifact Type Reference

When calling notebooklm_download_artifact, use these artifact_type values:

| artifact_type | Output format | |---------------|---------------| | audio | MP3/MP4 | | video | MP4 | | report | Markdown (.md) | | quiz | JSON | | flashcards | JSON | | infographic | PNG | | slide_deck | PDF | | data_table | CSV | | mind_map | JSON |

---

Development

git clone https://github.com/deji-adesoga/notebooklm-mcp
cd notebooklm-mcp
python -m venv .venv
.venv\Scripts\activate
pip install -e ".[dev]"

# Verify tools enumerate cleanly (no Google login needed)
python -c "from notebooklm_mcp.server import mcp; print([t.name for t in mcp.get_tools().values()])"

# Test with MCP Inspector
npx @modelcontextprotocol/inspector python -m notebooklm_mcp.server

---

License

MIT — see [LICENSE](LICENSE).

Underlying library: notebooklm-py (MIT).

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Ari-Agnt
• Source: Ari-Agnt/notebooklm-mcp
• License: MIT

Install and usage instructions live in the source repository linked above.