NotebookLM and Claude: A Powerful AI Workflow Combination

Claude has quickly become my favorite AI chatbot. For some reason, it took me so long to give it a real shot, but ever since I did, I’ve been using it practically exclusively for brainstorming, research, and, of course, coding. Although Claude is currently my go-to AI chatbot, the AI productivity tool you’ll always catch me obsessing over is NotebookLM.

I’ve penned hundreds of articles about it, used it to learn countless new things (programming, baking the perfect tiramisu, cramming for exams the night before), and I still keep finding new ways to make it work harder for me. One thing I absolutely love doing is pairing NotebookLM with other tools I’m already using. I’ve paired it with Notion, ChatGPT, Gemini, VS Code, etc. So, of course, it only made sense to try pairing it with Claude, and it’s quickly become one of my favorite combinations yet.

What is Claude?

Getting to know Claude

If you haven’t heard of Claude yet, it’s a conversational AI chatbot developed by Anthropic that’s quickly gained a response for being surprisingly human-like in its responses. Even before the launch of Claude Code, it was already known for handling programming and technical tasks better than most other LLMs. It also has a significantly larger context window compared to other AI tools, meaning it doesn’t forget instructions as often.

Claude also has unique features you won’t find in competitors, like Artifacts, that let you generate and interact with code or content in a live, editable environment instead of plain text. Claude currently has two paid tiers: Pro and Max. The former is priced at $17/month, while the latter costs $100/month. In addition to generating code, writing drafts, and searching the web in real-time, Claude also connects to a bunch of third-party tools, including Google Workspace (Gmail, Google Calendar, Google Docs, etc.) on the premium tiers.

Now, in addition to having a web-based interface and a mobile app, Claude also has a macOS and Windows app. I use the macOS app, and the reason why I’m mentioning this here will make sense later.

This setup lets me treat NotebookLM as an external hard drive

Just treat it like storage and let Claude do the work

A couple of days ago, I found out that Claude has a Projects feature that lets you create NotebookLM-like self-contained workspaces and knowledge bases, where you can upload documents and have hallucination-free chats with Claude. I tried to use the feature as a NotebookLM replacement, and it failed pretty badly because the file upload limit wasn’t very generous. I hit the cap after uploading just a handful of PDFs (on the Max plan, too)!

Since then, I’ve been trying to find a middleground that lets me take advantage of NotebookLM’s RAG capabilities and extended source limits, and Claude’s speed, reasoning, and coding strengths.

After a bit of digging, I found out I can use Anthropic’s Model Context Protocol (MCP) to connect Claude Desktop directly to NotebookLM. In simple terms, what this does is enable Claude to see my NotebookLM notebooks, interact with them and the sources within, and generate responses based on that data.

Related

I set up a daily log using NotebookLM and Google Calendar — here’s how

Hands down, the best thing I’ve done with NotebookLM so far.

The best part is I don’t need to so much as open NotebookLM myself once it’s set up. Because the MCP server handles the browser automation in the background, I can create new notebooks, add sources, and even trigger Audio Overviews without ever leaving the Claude Desktop app. It basically turns Claude into a “command center” for my research, keeping me in a single flow state instead of constantly tab-switching between my notes and my writing tool.

Here’s how I set it up, step by step

Just set it up once and forget about it

Setting this Claude + NotebookLM workflow up does take some manual work initially, but once it’s all set up, it’s smooth sailing. To begin, make sure you have the Claude Desktop app installed on your device. Then, if you’re on a Mac, run this in the terminal to download uv, a Python tool manager:

curl -LsSf https://astral.sh/uv/install.sh | sh

If it says “uv: command not found” after installing, run this command to refresh your shell:

source $HOME/.local/bin/env

On Windows, open PowerShell and run the following command:

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Then, you’ll need to download the bridge software. Paste the following command in your terminal on both macOS and Windows:

uv tool install notebooklm-mcp-server

You then need to authenticate the server with your Google account. To do so, run:

notebooklm-mcp-auth

Once you do, a Chrome window will appear and you’ll need to log into your Google account. After that, you need to create a specific configuration file to let Claude Desktop know the server exists. On Mac, you can simply run the command below to create and open the file:

mkdir -p ~/Library/Application Support/Claude/ && open -e ~/Library/Application Support/Claude/claude_desktop_config.json

On Windows, press Win + R and paste the following command:

%APPDATA%Claude

Then, create a new file named: claude_desktop_config.json

Finally, paste the following code into the file you just created (both on Mac/Windows) and save it:

{
"mcpServers": {
"notebooklm": {
"command": "uvx",
"args": [
"notebooklm-mcp-server"
]
}
}
}

Restart Claude Desktop, and you can confirm it’s working by asking Claude to list all your NotebookLM notebooks.

Now, you can simply type a prompt in a Claude chat asking it to create a new notebook, populate it with sources, retrieve information from the notebook, generate one of the Studio outputs — anything!

This is the closest you’ll get to the best of both worlds

With this setup, you effectively stay in one interface (i.e., Claude) for the most part, since it has direct access to NotebookLM, which acts as the knowledge base. Admittedly, this is a workaround, and it isn’t perfect, but it’s a solid way to take advantage of NotebookLM’s source-grounding capabilities and source limits alongside Claude’s superior reasoning and coding abilities.

Related

I made NotebookLM my personal assistant by pairing it with an agentic AI browser

The combo I’ve always needed.

As someone who has been using Claude extensively for coding projects and NotebookLM for documentation, this has been a massive time-saver. Instead of having to export my documentation from NotebookLM and paste it into Claude every time I need to refer to something, I can do it all within the Claude interface. For anyone who deals with heavy documentation or complex research, it’s well worth the five minutes of terminal tinkering.