Supercharge your AI Agent with Google's NotebookLM. Generate podcasts, perform deep research, and query your knowledge base—all from within OpenClaw.
This skill integrates the power of Google NotebookLM directly into your OpenClaw agent, enabling capabilities that standard LLMs cannot match:
| Feature | Description |
|---|---|
| 🎙️ Audio Overviews | Generate "Deep Dive" podcasts from any document or URL. |
| 🎓 Learning Quizzes | Turn YouTube videos or notes into study quizzes instantly. |
| 🧠 Deep Research | Autonomous web research agent that finds and summarizes 50+ sources. |
| 📚 RAG Engine | Query your specific documents with grounded answers (zero hallucinations). |
| ⚡ MCP Native | Built on the Model Context Protocol for seamless agent integration. |
- Manage Notebooks: Create, list, delete, and rename.
- Source Management: Add URLs, PDFs (local), Drive files, and pasted text.
- Studio Production: Create Audio Overviews, FAQs, Study Guides, and Briefing Docs.
- Note Taking: Create and manage notes within notebooks.
Clone this repo into your agent's skills directory. OpenClaw handles the rest.
cd ~/.openclaw/workspace/skills
git clone https://github.com/pjhwa/openclaw-notebooklm-skill.git notebooklmEnsure uv is installed for efficient Python package management.
macOS / Linux:
# macOS
brew install uv
# Linux
curl -LsSf https://astral.sh/uv/install.sh | shThe agent will automatically install notebooklm-mcp-cli when it first runs.
Robust & Secure: We use a file-based cookie method to ensure stability in headless environments (Linux servers, Docker containers).
- Open NotebookLM in your desktop browser.
- Open DevTools (F12) → Network tab.
- Refresh the page.
- Find a request to
notebooklm.google.com(e.g.,batchExecute). - Copy the entire
cookievalue from the Request Headers.
Save this string to a file on the machine running OpenClaw.
mkdir -p ~/.nlm
nano ~/.nlm/cookies.txt
# Paste the cookie string and save(If you have notebooklm-mcp-cli installed locally, you can also run nlm login --manual)
To enable the skill, register it with mcporter using the command appropriate for your OS.
Linux / macOS (Bash/Zsh):
# Register with Auth Injection (Bash)
npx -y mcporter --config ~/.mcporter_config.json config add notebooklm --stdio \
"bash -c 'export NOTEBOOKLM_COOKIES=\$(cat ~/.nlm/cookies.txt); notebooklm-mcp --transport stdio'"Windows (PowerShell):
# 1. Create directory
New-Item -ItemType Directory -Force -Path "$HOME\.nlm"
# 2. Save cookies (Paste your cookie string into this file)
notepad "$HOME\.nlm\cookies.txt"
# 3. Register with Auth Injection (PowerShell)
# Note: We use a simplified command for Windows that reads the file content via PowerShell
npx -y mcporter --config "$HOME\.mcporter_config.json" config add notebooklm --stdio "powershell -c \"`$env:NOTEBOOKLM_COOKIES = Get-Content '$HOME\.nlm\cookies.txt' -Raw; notebooklm-mcp --transport stdio\""Test your setup before unleashing the agent.
1. Check Connection
npx -y mcporter --config ~/.mcporter_config.json list notebooklmExpected: List of tools (notebook_list, studio_create, etc.)
2. Check Auth
npx -y mcporter --config ~/.mcporter_config.json call notebooklm.notebook_list --args "{}"Expected: JSON list of your notebooks.
3. Test Source & Polling
(Advanced): Add a test source with wait: true.
npx -y mcporter --config ~/.mcporter_config.json call notebooklm.source_add --args '{"notebook_id": "UUID", "source_type": "text", "text": "Test", "title": "Test", "wait": true}'Detailed guides for developers and power users:
- Architecture & Internals: Data flow, authentication logic, and SKILL.md structure.
- Advanced Scenarios: Workflows for "Podcast Generation", "Deep Research", and "RAG".
Contributions are welcome! Please open an issue or pull request.
MIT License