NotebookLM MCP

Overview

NotebookLM MCP (part of the notebooklm-mcp-cli package) is a production-ready Model Context Protocol (MCP) server and unified CLI (nlm) that gives any MCP-compatible AI agent full programmatic access to Google NotebookLM.

Built on undocumented NotebookLM APIs (with browser cookie auth), it turns NotebookLM into your personal zero-hallucination research engine that agents like Claude, Cursor, Gemini CLI, and Antigravity can query, update, and orchestrate directly.

Key Features

• 35+ MCP Tools exposed as natural-language functions:

  • notebook_list, notebook_create, notebook_query, notebook_share
  • source_add (URL, file, Google Drive, text)
  • studio_create / studio_revise (audio overviews, podcasts, video, slides)
  • research_start (web + Drive research with auto-import)
  • download_artifact, cross_notebook_query, batch operations, pipelines, tagging

• Full NotebookLM Automation:

  • Create notebooks & podcasts in one prompt
  • Add sources from anywhere
  • Generate, revise, and download studio content
  • Persistent multi-profile authentication

• CLI (nlm) for scripting and manual use:

  nlm notebook create "My Research"
  nlm source add MyResearch --url https://example.com
  nlm audio create MyResearch --confirm
  nlm research start "enterprise AI ROI" --import-top 10
  

• One-Command Setup for AI tools:

  nlm setup add claude-code
  nlm setup add cursor
  nlm setup add gemini
  

• Smart Extras: Skill guides (nlm skill install), doctor diagnostics, pipelines, cross-notebook queries, public/invite sharing.

Installation

# Recommended (uv)
uv tool install notebooklm-mcp-cli

# Or pip / pipx
pip install notebooklm-mcp-cli
pipx install notebooklm-mcp-cli

Then run nlm login (opens browser once) and nlm setup add claude-code (or your preferred client).

Use Cases

• Long-term memory for Claude/Cursor: Feed all your docs into NotebookLM once, then let your coding agent query it with perfect citations.
• Research + Content Pipelines: Start web research → auto-import sources → generate podcast → export slides.
• Multi-Agent Orchestration: n8n, CrewAI, AutoGen, and any MCP client can control NotebookLM via REST or stdio.
• Zero-Hallucination Knowledge Base: Agents refuse unknown answers and always cite your uploaded sources.

Why NotebookLM MCP?

NotebookLM (powered by Gemini 2.5) already synthesizes your documents better than local RAG. This MCP server removes the manual copy-paste step, letting your favorite AI agents talk to it directly — saving tokens, eliminating hallucinations, and unlocking full automation.

Supports free/Pro tiers (Enterprise untested). MIT licensed.

Quick Start for MCP Clients

1. Install & authenticate (nlm login)
2. Run the server: notebooklm-mcp
3. Connect in Claude Code / Cursor / Gemini CLI via the auto-setup command
4. Ask: "Create a notebook about quantum computing and generate a podcast from the latest arXiv papers."

Full documentation, tool reference, and CLI guide available in the repo.