Imported from akillness/jeo-skills (.agent-skills/notebooklm/SKILL.md). Install upstream with npx skills add akillness/jeo-skills --skill notebooklm. Copyright stays with the author (MIT).

notebooklm — Google NotebookLM Integration for Claude Code

Query your NotebookLM notebooks directly. Source-grounded, citation-backed answers without leaving the editor.

notebooklm brings Google NotebookLM into your Claude Code workflow via browser automation (Patchright). Instead of switching tabs and copy-pasting, you ask questions from the terminal and get grounded answers from your uploaded sources.

Installation

Plugin (Claude Code — recommended)

claude plugin marketplace add PleasePrompto/notebooklm-skill

Manual (git clone)

mkdir -p ~/.claude/skills
cd ~/.claude/skills
git clone https://github.com/PleasePrompto/notebooklm-skill.git notebooklm

Skill (via skills CLI)

npx skills add https://github.com/akillness/jeo-skills --skill notebooklm

Requirements

• Python 3.10+ — for browser automation scripts
• Chrome — installed via patchright install chrome (Chromium will NOT work)
• Google account — a dedicated account is recommended for automation
• Local Claude Code — web UI is not supported (sandbox restrictions)
• Internet access — required to reach notebooklm.google.com

Setup (first-time)

# 1. Install Python dependencies (auto-runs on first use)
cd ~/.claude/skills/notebooklm
python scripts/run.py auth_manager.py setup

# 2. A Chrome window opens — log in to Google manually
# The session is saved and reused for future queries

# 3. Add your first notebook
python scripts/run.py notebook_manager.py add \
  --url "https://notebooklm.google.com/notebook/YOUR_ID" \
  --name "my-research" \
  --description "Research notes on topic X"

When to use this skill

• You have documents uploaded to NotebookLM and need citation-backed answers during development
• You want to cross-reference your research notes while writing code without switching windows
• You need source-grounded analysis from a curated document set (not a general web search)
• You are doing research synthesis across multiple notebooks on related topics

Do not use when

• You need live web search — use scrapling or web-research instead
• You need general knowledge not in your uploaded documents — ask Claude directly
• You are running Claude from the web UI — browser automation is blocked in the sandbox
• Your Google account has exceeded the 50 queries/day free-tier limit

Usage

Ask a question

# Ask the currently active notebook
python scripts/run.py ask_question.py --question "What are the key findings about X?"

# Ask a specific notebook by ID
python scripts/run.py ask_question.py \
  --notebook-id "my-research" \
  --question "Summarize the methodology section"

# Ask with browser visible (useful for debugging)
python scripts/run.py ask_question.py \
  --question "What does source 3 say about Y?" \
  --show-browser

Manage notebooks

# List all notebooks
python scripts/run.py notebook_manager.py list

# Search notebooks
python scripts/run.py notebook_manager.py search "machine learning"

# Activate a notebook (sets it as the default)
python scripts/run.py notebook_manager.py activate --id "my-research"

# Remove a notebook from the library
python scripts/run.py notebook_manager.py remove --id "old-notes"

Authentication management

# Check auth status
python scripts/run.py auth_manager.py status

# Re-authenticate if session expired
python scripts/run.py auth_manager.py reauth

# Clear all auth data
python scripts/run.py auth_manager.py clear

Data storage

All user data is stored locally at ~/.claude/skills/notebooklm/data/:

data/
├── library.json          # Your notebook registry
├── auth_info.json        # Authentication metadata
├── browser_state/        # Persistent browser session
│   ├── browser_profile/  # Chrome profile data
│   └── state.json        # Cookies and localStorage
└── sessions.json         # Active session tracking

Key constraints

Troubleshooting

Full troubleshooting guide: ~/.claude/skills/notebooklm/references/troubleshooting.md

Follow-up protocol

After each answer, always assess completeness:

"Is that ALL you need to know from this notebook on this topic?"

NotebookLM can answer follow-up questions within the same session context. Ask targeted follow-ups before switching notebooks to get the most grounded coverage.

References

• PleasePrompto/notebooklm-skill — source repository
• ~/.claude/skills/notebooklm/references/usage_patterns.md — advanced usage patterns
• ~/.claude/skills/notebooklm/references/api_reference.md — programmatic usage
• ~/.claude/skills/notebooklm/references/troubleshooting.md — troubleshooting guide
• ~/.claude/skills/notebooklm/AUTHENTICATION.md — hybrid auth system details

Instructions

1. Identify the task trigger and expected output.
2. Follow the workflow steps in this skill from top to bottom.
3. Validate outputs before moving to the next step.
4. Capture blockers and fallback path if any step fails.

Examples

• Example: Apply this skill to a small scope first, then scale to full scope after validation passes.

Best practices

• Keep outputs deterministic and auditable.
• Prefer small reversible changes over broad risky edits.
• Record assumptions explicitly.