notebooklm-mcp-server

💡 What is this?

Google's Gemini Notebook (renamed from NotebookLM in July 2026) is the best grounded-answer engine on the market: it only answers from the sources you give it, with citations. This MCP server hands that power to your AI agents — Claude, Gemini, Antigravity, Cursor and any other MCP client — so they can build notebooks, run Deep Research, generate podcasts, videos, quizzes, flashcards, reports and mind maps, take notes, and share the results with your team. All programmatically, with zero hallucinations.

> [!NOTE]

> NotebookLM is now Gemini Notebook. Google renamed the product on July 16, 2026 (see notebook.google). Same product, same notebooks, same API surface — this server keeps working unchanged, and the npm package keeps its `notebooklm-mcp-server` name.

🚀 Installation

Option 1 — Global install (recommended)

npm install -g notebooklm-mcp-server

> [!NOTE]

> The server checks for updates on startup and keeps itself current automatically.

Option 2 — Zero-install with NPX

npx -y notebooklm-mcp-server auth   # authenticate
npx -y notebooklm-mcp-server start  # run the server

🔑 Authentication

1. Run the interactive login (opens a Chromium window):

npx notebooklm-mcp-server auth

2. Sign in with your Google account. When the notebook list appears, the session cookies (including Google's rotating `__Secure-1PSIDTS` token) are captured and saved to `~/.notebooklm-mcp/auth.json`.

3. Done — the server loads the session automatically and refreshes the rotating token by itself.

> [!TIP]

> If the session ever expires, run `npx notebooklm-mcp-server auth` again in a terminal, then call the `refresh_auth` MCP tool (or just restart your client) to pick up the new cookies without reconfiguring anything.

⚡ Connect Your AI Client

🤖 Claude Code

claude mcp add notebooklm -- npx -y notebooklm-mcp-server start

💬 Claude Desktop

Add to `claude_desktop_config.json` (Settings → Developer → Edit Config):

{
  "mcpServers": {
    "notebooklm": {
      "command": "npx",
      "args": ["-y", "notebooklm-mcp-server", "start"]
    }
  }
}

🌌 Antigravity CLI

Antigravity's CLI manages MCP servers through a JSON config file:

• Global: `~/.gemini/config/mcp_config.json`
• Per-workspace: `.agents/mcp_config.json` in your project

{
  "mcpServers": {
    "notebooklm": {
      "command": "npx",
      "args": ["-y", "notebooklm-mcp-server", "start"]
    }
  }
}

Then type `/mcp` in the Antigravity prompt panel to open the MCP Manager: check the connection status ring, reload the config, or inspect logs — no restart needed. Your agents can now cite real sources in every mission.

🖥️ Antigravity IDE

The IDE reads the same config files as the CLI (`~/.gemini/config/mcp_config.json` or `.agents/mcp_config.json`). Alternatively, open the MCP Store panel, choose *Add custom server*, and paste the same JSON snippet. Reload the server list and the NotebookLM tools appear in the agent's toolbox.

💎 Gemini CLI

gemini mcp add notebooklm --scope user -- npx -y notebooklm-mcp-server start

⌨️ Cursor

Add to `.cursor/mcp.json` in your project (or `~/.cursor/mcp.json` globally):

{
  "mcpServers": {
    "notebooklm": {
      "command": "npx",
      "args": ["-y", "notebooklm-mcp-server", "start"]
    }
  }
}

🏄 Windsurf

Add the same `mcpServers` block to `~/.codeium/windsurf/mcp_config.json`.

🧩 VS Code (Cline)

In Cline: MCP Servers → Configure and add the same `mcpServers` block shown above.

📖 Tool Reference

47 tools, grouped by what they do. Full parameter signatures live in docs/TOOLS.md.

📒 Notebooks

🔗 Sharing

🖇️ Sources

💬 Chat

🔍 Research

🎨 Studio

📦 Artifacts

📝 Notes & Mind Maps

⚙️ System

🧪 Recipes

Real prompts you can paste into any connected agent — it picks the right tools by itself.

🎙️ From research question to podcast

> *"Research the latest on solid-state batteries with deep research, build a notebook called 'Solid State 2026' with the best sources, and generate a brief audio overview in English. Give me the link when it's done."*

The agent chains `notebook_create` → `research_start(mode: deep)` → `research_poll` → `research_import` → `audio_overview_create(format: brief)` → `studio_poll`, and returns the notebook URL with the finished podcast.

🎓 Study kit from your PDFs

> *"Take the three PDFs in ./lectures, put them in a new notebook, and make me a hard quiz plus flashcards on the key concepts. Show me the quiz questions here."*

The agent runs `notebook_add_local_file` per PDF, then `quiz_create(difficulty: hard)` + `flashcards_create`, polls with `studio_poll`, and pulls the questions out with `artifact_content_get`.

👥 Team briefing, shared

> *"Build a briefing notebook from these five links about our competitor, write a report, export it to Google Docs, and share the notebook with ana@example.com as a viewer."*

The agent chains `notebook_add_url` ×5 → `report_create` → `studio_poll` → `artifact_export(format: docs)` → `notebook_share(user_email, user_role: viewer)` — and your teammate finds the notebook and the Doc waiting.

🛠️ Troubleshooting

🧑‍💻 Development

git clone https://github.com/moodRobotics/notebooklm-mcp-server.git
cd notebooklm-mcp-server
npm install
npm run build
npm run typecheck

Architecture notes live in docs/ARCHITECTURE.md, the roadmap in docs/ROADMAP.md.

🌐 Localization

This README is available in English, Español, Français, Português and Deutsch. If you edit one, please keep the others in sync (`npm run docs:check`).

📄 License

MIT — Developed with ❤️ by moodRobotics.

> [!IMPORTANT]

> This is an unofficial community project. It is not affiliated with or endorsed by Google. It relies on reverse-engineered internal APIs that may change without notice; use it with a Google account you are comfortable automating.