NotebookLM MCP (khengyun)

A no-browser take on NotebookLM automation — talks to an RPC backend instead of driving Chrome, and can merge sources across notebooks or generate mind maps.

License: MIT · Best for: Developers · Anyone merging sources across multiple notebooks

What it is

A Python-based MCP server for NotebookLM built on an RPC client (notebooklm-py) rather than browser automation — there's no Chrome process to keep running at runtime. Its 28 tools cover chat, notebook and source management, plus two things the other servers in this list don't: merging sources from multiple notebooks into a new one, and generating and reading mind maps.

Why NotebookLM users should care

The no-browser architecture is the headline difference: lighter to run, nothing to keep alive between calls, at the cost of the DOM-level citation scraping the browser-driven alternatives support more directly. If your workflow is less about pulling citations and more about consolidating scattered notebooks — merging sources from three different notebooks into one, say, or reading back a mind map programmatically — this is the only one of the three built for that.

Key capabilities

Installation summary

Requires Python 3.10+. Install it, run a one-time login, then initialize it against a notebook URL to write a local config file before starting the server. There's no npm package for this one — it's a Python project distributed on PyPI, not a Node one.

uv add notebooklm-mcp && uv run notebooklm login

Official links

FAQ

Why no Chrome/browser requirement?

It talks to NotebookLM through notebooklm-py, an RPC client, instead of automating a real browser session — lighter to run, but citation extraction isn't as deep as the browser-driven alternatives.

Is there an npm package?

No — this is a Python project distributed on PyPI, installed with uv or pip, not npm.

What's the mind map feature for?

NotebookLM can generate a visual mind map of a notebook's sources; this server exposes tools to generate, list, and read those mind maps programmatically, which neither of the other two servers here currently supports.

Is it actively maintained?

Yes — it has the smallest GitHub following of the three, but has shipped updates as recently as mid-2026.

← Back to MCP Servers · Resource Library