rag-retrieval / mcp
NotebookLM MCP Server
Capability: NotebookLM MCP Server
Use it when
- Your agent needs cited, grounded answers from Google NotebookLM notebooks but NotebookLM offers no public API
- You want to automate NotebookLM workflows like adding sources or generating Audio Overviews from Claude Code, Cursor, or an n8n HTTP pipeline
- You maintain a library of research notebooks and want agents to search, select, and query them with DOM-level citation extraction
What it solves
Not the fit when
- Official or vendor-supported NotebookLM API integration; this project automates the web product through a real browser
- Encrypted credential storage; account separation is by persistent Chrome profile directory
- First-time authentication on a headless Linux host without a display or xvfb-run
- Publicly exposing Streamable HTTP mode without an external authentication, TLS, and network-access layer
- Treating NotebookLM answers, cited documents, or embedded instructions as trusted; retain the default AI-generated/untrusted marker
- general web browsing automation
- building your own vector-store RAG pipeline
- bulk Google Drive document management
Install
claude mcp add notebooklm -- npx notebooklm-mcp@latest (requires Node.js >= 18 and Chrome)
Invoke
Run the setup_auth tool once to log into Google in a visible Chrome, then call ask_question against a selected notebook (source_format=json for structured citations), add_source to ingest URLs or text, and generate_audio/download_audio for Audio Overviews
Alternatives
No reviewed alternatives recorded yet.