mcp-notebooklm

# MCP NotebookLM Deep research and analysis via NotebookLM MCP (lazy-loaded via ToolSearch). ## Loading ``` ToolSearch: +notebooklm ``` ## MCP Tools | Task | Tool | |------|------| | Create notebook | `mcp__notebooklm__add_notebook` | | Select notebook | `mcp__notebooklm__select_notebook` | | Get details | `mcp__notebooklm__get_notebook` | | Ask question | `mcp__notebooklm__ask_question` | | List notebooks | `mcp__notebooklm__list_notebooks` | | Search notebooks | `mcp__notebooklm__search_notebooks` | | Update sources | `mcp__notebooklm__update_notebook` | | Remove notebook | `mcp__notebooklm__remove_notebook` | | Cleanup data | `mcp__notebooklm__cleanup_data` | | Health check | `mcp__notebooklm__get_health` | | Session mgmt | `list_sessions`, `close_session`, `reset_session` | | Auth | `mcp__notebooklm__setup_auth`, `mcp__notebooklm__re_auth` | ## Standard Workflow ### 1. Create or Reuse Notebook ``` list_notebooks -> check for existing match add_notebook(title="Topic Name", sources=[url1, url2, ...]) ``` Sources: YouTube URLs (`https://youtube.com/watch?v=...`), web URLs, or raw text. Up to 50 per notebook. ### 2. Extract Insights Ask specific, focused questions: ``` ask_question("What are the key themes across these sources?") ask_question("What topics get the most engagement and why?") ask_question("What content gaps exist that aren't covered?") ask_question("What are the main takeaways for someone new to this?") ``` ### 3. Request Deliverables (if needed) Use `ask_question` with clear requests: - **Audio overview**: "Generate an audio overview / podcast summary" - **Infographic**: "Create an infographic of key findings" - **Mind map**: "Create a mind map of main concepts" - **Study guide**: "Create a study guide with key concepts and questions" - **FAQ**: "Generate a FAQ based on these sources" - **Briefing doc**: "Create a briefing document of the most important points" Visual deliverables (infographic, mind map) can take 5-15 minutes. ### 4. Synthesize and Return Summarize analysis clearly. Include any links or content NotebookLM returned. ## Best Practices - **One notebook per research topic** -- keeps analysis focused - **Diverse sources** -- mix video, text, docs for balanced perspective - **Specific questions** -- "What pricing strategies do top channels use?" beats "Tell me about the videos" - **Search before creating** -- reuse existing notebooks when possible - **Clean up** -- use `cleanup_data` to remove stale notebooks periodically ## Auth Issues If NotebookLM returns auth errors: 1. Try `mcp__notebooklm__re_auth` first 2. If that fails, `mcp__notebooklm__setup_auth` for fresh setup 3. Check `mcp__notebooklm__get_health` for service status