NotebookLM

This MCP server lets your CLI assistants (like Claude, Cursor, Codex) talk directly to NotebookLM to fetch expert, zero-illusion answers from your own notes. It enables your agent to query NotebookLM and receive integrated, source-backed guidance that informs code and decisions without manual copy-paste between editor and knowledge base.

How to use

You connect your MCP client to the NotebookLM MCP server, then let your agent query NotebookLM for guidance, code examples, and contextual knowledge drawn from your notebooks. The agent will build a sequence of informed questions before coding to ensure it fully understands the task and references the right sources.

Key usage patterns include selecting a notebook as the current knowledge source, prompting the agent with your intent, and letting it fetch, synthesize, and present concrete implementation steps. You can attach multiple notebooks to a shared library with tags and descriptions; the agent will automatically pick the most relevant notebook for the current task.

How to install

Prerequisites: you need Node.js and npm installed on your machine. Ensure you have a modern runtime (Node.js 14+ is typically suitable). Then follow the concrete steps below to install and configure the MCP server and connect your MCP client.

1. Install the MCP server for NotebookLM using the following command from your terminal. This registers the MCP with your CLI agent.

claude mcp add notebooklm npx notebooklm-mcp-zh@latest

2. If you are using Cursor, add a local MCP configuration entry that points to the runner. Create or edit the MCP configuration file to include the following entry.

{
  "mcpServers": {
    "notebooklm": {
      "command": "npx",
      "args": ["-y", "notebooklm-mcp-zh@latest"]
    }
  }
}

Additional setup steps

1. Verify access to NotebookLM by logging in through the connected browser when prompted by your agent. This step establishes your session.

2. Build your knowledge library in NotebookLM by uploading PDFs, documents, websites, repositories, or videos. Add descriptive tags to help your agent locate the most relevant notebook for each task.

3. Start using the agent to request NotebookLM insights. You can share NotebookLM links with your library and reference them via tagged queries.

Notes on usage

The agent will avoid guessing information not present in your notebooks. If a detail is unavailable, it will refrain from fabricating an API or other elements. You can guide the agent to focus on specific notebooks or tags to refine results.

If you need to reset the session or change knowledge sources, use the library management commands to switch notebooks or re-run the cleanup process to refresh available data.

Available tools

ask_question

Ask a targeted question to narrow down requirements or confirm assumptions before coding.

get_health

Check the health/status of the NotebookLM MCP connection and library access.

list_notebooks

List all notebooks currently stored in your NotebookLM library.

select_notebook

Choose which notebook is the active context for subsequent queries.

get_notebook

Fetch the contents or metadata of a specific notebook.

setup_auth

Prepare authentication flow for NotebookLM access within the MCP workflow.

list_sessions

List active sessions or connections associated with NotebookLM in the MCP environment.

add_notebook

Add a new notebook to your NotebookLM library from a link or local upload.

update_notebook

Update metadata or sources for an existing notebook.

search_notebooks

Search across notebooks to find relevant sources and references.

cleanup_data

Remove or reorganize library data to maintain a clean knowledge base.

re_auth

Re-run authentication to refresh access when credentials expire.