Managing NotebookLM

This skill provides a comprehensive interface for interacting with Google NotebookLM through the Model Context Protocol (MCP). It enables deep research, content generation, and knowledge base management.

When to use this skill

• Deep Research: When the user needs to investigate a topic thoroughly using web search or Drive documents.
• Content Generation: When the user wants to turn notes into Audio Overviews, Video Overviews, Study Guides, FAQs, or Briefing Docs.
• Knowledge Management: When the user needs to organize, query, or synthesize information from multiple sources (PDFs, URLs, etc.).
• Notebook Administration: When creating, renaming, or cleaning up notebooks and sources.

Workflows

1. Deep Research Flow (Finding New Information)

Use this flow when the user wants to learn about a new topic or find new sources.

1. Start Research: Use research_start with mode="deep" (web only) or mode="fast".
   • Tip: Specify notebook_id if adding to an existing notebook, otherwise a new one is created.
2. Poll Status: Loop research_status until status="completed".
   • Note: This tool blocks by default up to max_wait seconds, so a single call might be enough.
3. Import Sources: Use research_import to add the discovered sources to the notebook.
   • Tip: You can filter sources using source_indices if the research found irrelevant items.

2. Knowledge Query Flow (Q&A on Existing Sources)

Use this flow when the user asks questions about documents already added to a notebook.

1. List Notebooks (Optional): Use notebook_list to find the correct notebook_id.
2. Verify Sources (Optional): Use notebook_get or source_list_drive to ensure the relevant documents are present.
   • Tip: Use source_sync_drive if Drive documents might be stale.
3. Query: Use notebook_query to ask specific questions.
   • Note: This uses RAG (Retrieval-Augmented Generation) on the notebook's sources.

3. Content Studio Flow (Generating Artifacts)

Use this flow to create consumable media from the notebook's content.

1. Select Format: Choose the appropriate tool based on user request:
   • Audio/Podcasts: audio_overview_create (formats: deep_dive, brief, etc.)
   • Video: video_overview_create (formats: explainer, brief; styles: whiteboard, anime, etc.)
   • Visuals: infographic_create, slide_deck_create, mind_map_create
   • Study Aids: flashcards_create, quiz_create, data_table_create
   • Documents: report_create
2. Confirm: All creation tools require confirm=True after explicit user approval.
3. Check Status: Use studio_status to retrieve the URLs of generated assets (Audio/Video).

4. Authentication Management

The primary authentication source is the file /Users/ccachero/.notebooklm-mcp/auth.json.

When encountering 401/403 errors or initial setup:

1. Load Credentials: View the content of /Users/ccachero/.notebooklm-mcp/auth.json.
2. Inject Session: Use the save_auth_tokens tool.
   • Map JSON cookie -> tool arg cookies
   • Map JSON csrf_token (or X-Chrzd) -> tool arg csrf_token
3. Refresh: Call refresh_auth to finalize the session restore.
4. Emergency Fallback: If the file is missing or checks fail, ask the user to run notebooklm-mcp-auth in their terminal to regenerate the JSON.

Tool Reference Guide

Core Management

• notebook_create(title): Create a new workspace.
• notebook_list(): See all available notebooks.
• notebook_delete(id, confirm): Remove a notebook (Irreversible).
• notebook_rename(id, title): Change title.
• chat_configure(id, goal, prompt): Customize the system prompt/persona for the notebook chat.

Source Management

• notebook_add_url(id, url): Add web pages or YouTube videos.
• notebook_add_text(id, text): Add raw text / copy-paste.
• notebook_add_drive(id, doc_id): Add Google Docs/Slides/PDFs from Drive.
• source_get_content(id): Retrieve raw text of a specific source.
• source_delete(id): Remove a source.

Advanced Features

• research_start: The entry point for the "Deep Research" agentic capability.
• studio_status: The dashboard for checking generation progress of Audio/Video.