NotebookLM Research Assistant

Overview

Query your Google NotebookLM notebooks directly from Claude Code. Get source-grounded, citation-backed answers from Gemini based on your uploaded documents. Each question opens a fresh browser session, drastically reducing hallucinations through document-only responses.

When to Use

• Need document-grounded answers (reducing AI hallucination)
• Have reference materials uploaded to NotebookLM
• Want citation-backed responses (each answer references source paragraphs)
• User mentions NotebookLM or shares a NotebookLM URL

Workflow

1. Check auth — python scripts/run.py auth_manager.py status
2. Authenticate — python scripts/run.py auth_manager.py setup (browser visible, manual Google login)
3. Manage notebooks — Add, list, search, activate notebooks
4. Ask questions — python scripts/run.py ask_question.py --question "..."
5. Follow-up loop — Analyze completeness, ask follow-ups if gaps exist
6. Synthesize — Combine all answers before responding to user

Core Commands

Authentication

python scripts/run.py auth_manager.py setup    # First-time setup (browser visible)
python scripts/run.py auth_manager.py status   # Check status
python scripts/run.py auth_manager.py reauth   # Re-authenticate

Notebook Management

python scripts/run.py notebook_manager.py list
python scripts/run.py notebook_manager.py add --url URL --name "Name" --description "Description" --topics "topic1,topic2"
python scripts/run.py notebook_manager.py search --query "keyword"
python scripts/run.py notebook_manager.py activate --id ID

Asking Questions

python scripts/run.py ask_question.py --question "Your question"
python scripts/run.py ask_question.py --question "..." --notebook-id ID
python scripts/run.py ask_question.py --question "..." --show-browser

Smart Add (When User Doesn't Provide Details)

# Step 1: Query notebook to discover content
python scripts/run.py ask_question.py --question "What is the content of this notebook?" --notebook-url "[URL]"

# Step 2: Add with discovered metadata
python scripts/run.py notebook_manager.py add --url "[URL]" --name "Discovered Name" --description "Discovered Description" --topics "discovered,topics"

Follow-Up Mechanism

After every NotebookLM answer:

1. STOP — Don't immediately respond to user
2. ANALYZE — Compare answer to user's original request
3. IDENTIFY GAPS — More information needed?
4. ASK FOLLOW-UP — If gaps exist, query again with context
5. SYNTHESIZE — Combine all answers before final response

Limitations

Troubleshooting

Key Rules

• Always use run.py — Manages virtualenv and dependencies automatically
• Check auth first — Before any operation
• Never guess notebook descriptions — Use Smart Add to discover content
• Follow up aggressively — First answer is rarely complete