Connecting Claude and NotebookLM via MCP | Setup Guide and Research Workflow

AI Chat Article Summarypowered by Claude

Google's NotebookLM and Anthropic's Claude are both leading AI tools of 2026. Until now, the common approach was to use them separately — NotebookLM for organizing research materials and Claude for writing and reasoning. However, with NotebookLM's adoption of MCP (Model Context Protocol) in early 2026, you can now control NotebookLM directly from Claude Desktop.

Without switching between tabs, you can create NotebookLM notebooks, add sources, and run queries all within the flow of a Claude conversation — this integration has a significant impact on research workflows and is rapidly spreading among researchers and content creators worldwide (source: XDA Developers, 2026-02-23).

This article explains the MCP connection setup for both Mac and Windows, and introduces practical research workflows.

結論powered by Claude
Google's NotebookLM and Anthropic's Claude are both leading AI tools of 2026. Until now, the common approach was to use them separately — NotebookLM for organizing research materials and Claude for writing and reasoning. However, with NotebookLM's adoption of MCP (Model Context Protocol) in early 2026, you can now control NotebookLM directly from Claude Desktop.
目次 (17)

Understanding the Strengths of Claude and NotebookLM

Before diving into setup, let's clarify why combining these two tools makes sense.

NotebookLM's Strengths

  • Strictly references uploaded sources (PDFs, YouTube, URLs, etc.) and does not introduce information outside those sources
  • Easy to verify citations since every reference is explicitly quoted
  • Offline learning via Audio Overview (audio summaries)

Claude's Strengths

  • Excels at reasoning, comparison, and synthesis across sources
  • Long-form transcription, structuring, and reformatting
  • Executes complex instructions while maintaining conversational context

Connecting the two via MCP enables a clear division of roles: "NotebookLM maintains the knowledge base for research, while Claude handles analysis and writing" (source: Medium — NotebookLM + Claude via MCP).

Prerequisites for MCP Connection

Before getting started, make sure you have the following:

  • Claude Desktop app (install the latest version from anthropic.com)
  • Google account (with access to NotebookLM)
  • uv (Python package manager)

uv is a tool that automatically manages Python environments without polluting your existing Python setup. If you haven't installed it yet, follow the steps in the next section.

Setup Steps (Mac/Windows)

Step 1: Install uv

For Mac (run in Terminal):

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

After installation, restart your terminal or run source ~/.zshrc (or ~/.bashrc) to apply the path changes.

For Windows (run in PowerShell):

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

Step 2: Install the NotebookLM MCP Server

uv tool install notebooklm-mcp-server

If the package name has changed, try notebooklm-mcp-cli:

uv tool install notebooklm-mcp-cli

Step 3: Complete Google Authentication

notebooklm-mcp-auth

A browser window will open displaying the Google account authentication screen. Authorize using the Google account you use with NotebookLM.

Step 4: Edit the Claude Desktop Configuration File

For Mac, run the following in Terminal to open the configuration file:

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

For Windows, press Win + R, type %APPDATA%\Claude to open the folder, then open claude_desktop_config.json with a text editor.

Step 5: Write the MCP Server Configuration

Add the following content to claude_desktop_config.json (if existing configuration is present, append within the mcpServers object):

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

If you installed notebooklm-mcp-cli, change args as follows:

"args": ["--from", "notebooklm-mcp-cli", "notebooklm-mcp"]

Step 6: Restart Claude Desktop

Changes to the configuration file take effect after a restart. Fully quit and restart Claude Desktop. If the NotebookLM icon appears in the tools panel, the connection is successful.

What You Can Do from Claude After Connecting

Once the MCP connection is established, you can perform the following operations directly from the Claude chat interface (source: XDA Developers):

  1. Create new notebooks — Simply say "Create a notebook on this topic" and a notebook will be generated in NotebookLM
  2. Add sources — Directly add URLs or search results to a NotebookLM notebook
  3. Run queries on notebooks — Claude sends questions to a specific notebook and retrieves answers with citations
  4. Trigger Audio Overview generation — Start generating audio summaries via Claude

These operations are most effective when combined with Claude's reasoning and writing capabilities. For example, an instruction like "Write a 1,000-word summary based on the content of this notebook" can be completed in a single conversation.

Practical Research Workflow Examples

Academic Paper Review

  1. Add the PDF or URL of the paper you want to read to a NotebookLM notebook (operable via Claude)
  2. Ask Claude to "Extract the research objectives, methodology, and conclusions from this notebook"
  3. Claude outputs a cited summary while referencing NotebookLM's sources
  4. Follow up with "Create a comparison table with related prior research" to leverage Claude's comparative analysis capabilities

Market Research for Content Creation

  1. Compile URLs of competitor articles and industry reports in NotebookLM
  2. Ask Claude to "List three differentiating points based on the information in this notebook"
  3. NotebookLM ensures source attribution while Claude reformats the content into reader-friendly prose

Organizing Meeting Minutes

  1. Add meeting transcript text as a source in NotebookLM
  2. Extract action items and decisions via Claude, then format them into an email

Important Notes and Common Issues

Allow Plenty of Time for Setup

While articles suggest setup is possible in "10 minutes," the process can take 30 minutes or more due to the Google authentication steps or package name changes. If you're trying this for the first time, set aside enough time.

Be Aware of Package Name Changes

As of 2026, a package name change from notebooklm-mcp-server to notebooklm-mcp-cli has been reported. If installation fails, try both names.

Manual Notebook Addition May Be Required

For formats that cannot be added as sources via MCP (such as certain local files), you may need to add them manually through the NotebookLM web interface.

Check Your Claude Desktop Version

The MCP feature is available in relatively recent versions of Claude Desktop. If the connection fails, update the app to the latest version.

When to Use NotebookLM Alone vs. the Integration

This integration is most effective when you need to "analyze and process information across multiple sources." On the other hand, if you simply want to "look up the contents of uploaded materials faithfully" or "verify source citations directly," using NotebookLM alone is simpler and more reliable.

For a comparison of Claude and NotebookLM (features, pricing, and usage guidelines), see "NotebookLM vs Claude Comparison | Choosing Between Research/Summary and Document Creation". We recommend reading that before starting with MCP integration.

Summary

With NotebookLM's MCP support, you can now control Google's research AI directly from Claude Desktop. Setup begins with installing uv and completes in 6 steps. Once connected, you can produce reliably cited outputs by combining NotebookLM's source management capabilities with Claude's reasoning and writing abilities.

The integration is especially effective in situations where source credibility is important — such as paper reviews, market research, and meeting minute organization. Start by testing the connection with a single trial notebook.

参考になったら ♡
Clauder Navi 編集部
@clauder_navi

Anthropic の Claude / Claude Code を中心に、日本のエンジニア向けに最新動向と実務 を毎日発信。 運営方針 は メディアについて をご覧ください。