Turn NotebookLM into a CLI and MCP Server for AI Agents
NotebookLM is great for summarizing documents, answering questions from your notes, and organizing research. But what if you could access all that from your terminal or hook it into an AI agent pipeline?
That’s exactly what this project does. It wraps NotebookLM’s functionality into a simple command-line tool and a Model Context Protocol (MCP) server, letting you query your notebooks programmatically. No browser, no clicking around.
What It Does
The repository at github.com/jacob-bd/notebooklm-mcp provides two main interfaces:
- A CLI – run queries against your NotebookLM notebooks directly from the terminal. Think
notebooklm ask "What were the key findings in my Q3 report?" - An MCP server – exposes your notebooks as a context provider compatible with MCP-based tools (like Anthropic’s Claude + MCP integrations, or custom agent frameworks). This means an AI agent can fetch relevant notebook content on the fly.
Under the hood, it authenticates with your Google account and reads notebook data via NotebookLM’s internal API. You get structured answers – not just raw text dumps – so it’s easy to pipe into scripts or other tools.
Why It’s Cool
No more copy-paste. If you keep research or project notes in NotebookLM, you’ve probably had to manually copy summaries into your CLI or agent context. This removes that friction.
Agent-friendly. The MCP integration is the real winner here. If you’re building agents that need to reference specific documents or notes (e.g., personal knowledge bases, code documentation, financial reports), they can now pull from NotebookLM without you writing custom scrapers.
Simple auth flow. It uses your existing Google session – no OAuth gymnastics. Just follow the setup steps and you’re good.
Lightweight. It’s a single Python package. You don’t need to spin up a service or manage a database.
How to Try It
-
Clone the repo:
git clone https://github.com/jacob-bd/notebooklm-mcp -
Install dependencies (Python 3.10+):
cd notebooklm-mcp && pip install -r requirements.txt -
Set up authentication – the README has clear step-by-step instructions for getting your Google credentials.
-
Run the CLI:
python -m notebooklm_mcp ask "Summarize my meeting notes from today"
Or fire up the MCP server:
python -m notebooklm_mcp serve
Full details, including how to configure it for Claude or custom agents, are in the repo’s README. It’s well written – no guesswork.
Final Thoughts
This is one of those “why didn’t I think of that” tools. It’s small, focused, and solves a real annoyance if you use NotebookLM regularly. For developers building AI agents that need to answer questions from personal notes or document collections, this saves hours of glue code.
Give it a spin if you’ve got notebooks sitting there with untapped potential. Your terminal will thank you.
Follow @githubprojects for more tools like this.