Loading source
Pulling the file list, source metadata, and syntax-aware rendering for this listing.
Source from repo
Query Google NotebookLM notebooks from Claude Code for source-grounded, citation-backed answers from Gemini.
Files
Skill
Size
Entrypoint
Format
Open file
Syntax-highlighted preview of this file as included in the skill package.
README.md
1<div align="center">23# NotebookLM Claude Code Skill45**Let [Claude Code](https://github.com/anthropics/claude-code) chat directly with NotebookLM for source-grounded answers based exclusively on your uploaded documents**67[](https://www.python.org/)8[](https://www.anthropic.com/news/skills)9[](https://github.com/PleasePrompto/notebooklm-mcp)10[](https://github.com/PleasePrompto/notebooklm-skill)1112> Use this skill to query your Google NotebookLM notebooks directly from Claude Code for source-grounded, citation-backed answers from Gemini. Browser automation, library management, persistent auth. Drastically reduced hallucinations - answers only from your uploaded documents.1314[Installation](#installation) • [Quick Start](#quick-start) • [Why NotebookLM](#why-notebooklm-not-local-rag) • [How It Works](#how-it-works) • [MCP Alternative](https://github.com/PleasePrompto/notebooklm-mcp)1516</div>1718---1920## ⚠️ Important: Local Claude Code Only2122**This skill works ONLY with local [Claude Code](https://github.com/anthropics/claude-code) installations, NOT in the web UI.**2324The web UI runs skills in a sandbox without network access, which this skill requires for browser automation. You must use [Claude Code](https://github.com/anthropics/claude-code) locally on your machine.2526---2728## The Problem2930When you tell [Claude Code](https://github.com/anthropics/claude-code) to "search through my local documentation", here's what happens:31- **Massive token consumption**: Searching through documentation means reading multiple files repeatedly32- **Inaccurate retrieval**: Searches for keywords, misses context and connections between docs33- **Hallucinations**: When it can't find something, it invents plausible-sounding APIs34- **Manual copy-paste**: Switching between NotebookLM browser and your editor constantly3536## The Solution3738This Claude Code Skill lets [Claude Code](https://github.com/anthropics/claude-code) chat directly with [**NotebookLM**](https://notebooklm.google/) — Google's **source-grounded knowledge base** powered by Gemini 2.5 that provides intelligent, synthesized answers exclusively from your uploaded documents.3940```41Your Task → Claude asks NotebookLM → Gemini synthesizes answer → Claude writes correct code42```4344**No more copy-paste dance**: Claude asks questions directly and gets answers straight back in the CLI. It builds deep understanding through automatic follow-ups, getting specific implementation details, edge cases, and best practices.4546---4748## Why NotebookLM, Not Local RAG?4950| Approach | Token Cost | Setup Time | Hallucinations | Answer Quality |51|----------|------------|------------|----------------|----------------|52| **Feed docs to Claude** | 🔴 Very high (multiple file reads) | Instant | Yes - fills gaps | Variable retrieval |53| **Web search** | 🟡 Medium | Instant | High - unreliable sources | Hit or miss |54| **Local RAG** | 🟡 Medium-High | Hours (embeddings, chunking) | Medium - retrieval gaps | Depends on setup |55| **NotebookLM Skill** | 🟢 Minimal | 5 minutes | **Minimal** - source-grounded only | Expert synthesis |5657### What Makes NotebookLM Superior?58591. **Pre-processed by Gemini**: Upload docs once, get instant expert knowledge602. **Natural language Q&A**: Not just retrieval — actual understanding and synthesis613. **Multi-source correlation**: Connects information across 50+ documents624. **Citation-backed**: Every answer includes source references635. **No infrastructure**: No vector DBs, embeddings, or chunking strategies needed6465---6667## Installation6869### The simplest installation ever:7071```bash72# 1. Create skills directory (if it doesn't exist)73mkdir -p ~/.claude/skills7475# 2. Clone this repository76cd ~/.claude/skills77git clone https://github.com/PleasePrompto/notebooklm-skill notebooklm7879# 3. That's it! Open Claude Code and say:80"What are my skills?"81```8283When you first use the skill, it automatically:84- Creates an isolated Python environment (`.venv`)85- Installs all dependencies including **Google Chrome**86- Sets up browser automation with Chrome (not Chromium) for maximum reliability87- Everything stays contained in the skill folder8889**Note:** The setup uses real Chrome instead of Chromium for cross-platform reliability, consistent browser fingerprinting, and better anti-detection with Google services9091---9293## Quick Start9495### 1. Check your skills9697Say in Claude Code:98```99"What skills do I have?"100```101102Claude will list your available skills including NotebookLM.103104### 2. Authenticate with Google (one-time)105106```107"Set up NotebookLM authentication"108```109*A Chrome window opens → log in with your Google account*110111### 3. Create your knowledge base112113Go to [notebooklm.google.com](https://notebooklm.google.com) → Create notebook → Upload your docs:114- 📄 PDFs, Google Docs, markdown files115- 🔗 Websites, GitHub repos116- 🎥 YouTube videos117- 📚 Multiple sources per notebook118119Share: **⚙️ Share → Anyone with link → Copy**120121### 4. Add to your library122123**Option A: Let Claude figure it out (Smart Add)**124```125"Query this notebook about its content and add it to my library: [your-link]"126```127Claude will automatically query the notebook to discover its content, then add it with appropriate metadata.128129**Option B: Manual add**130```131"Add this NotebookLM to my library: [your-link]"132```133Claude will ask for a name and topics, then save it for future use.134135### 5. Start researching136137```138"What does my React docs say about hooks?"139```140141Claude automatically selects the right notebook and gets the answer directly from NotebookLM.142143---144145## How It Works146147This is a **Claude Code Skill** - a local folder containing instructions and scripts that Claude Code can use when needed. Unlike the [MCP server version](https://github.com/PleasePrompto/notebooklm-mcp), this runs directly in Claude Code without needing a separate server.148149### Key Differences from MCP Server150151| Feature | This Skill | MCP Server |152|---------|------------|------------|153| **Protocol** | Claude Skills | Model Context Protocol |154| **Installation** | Clone to `~/.claude/skills` | `claude mcp add ...` |155| **Sessions** | Fresh browser each question | Persistent chat sessions |156| **Compatibility** | Claude Code only (local) | Claude Code, Codex, Cursor, etc. |157| **Language** | Python | TypeScript |158| **Distribution** | Git clone | npm package |159160### Architecture161162```163~/.claude/skills/notebooklm/164├── SKILL.md # Instructions for Claude165├── scripts/ # Python automation scripts166│ ├── ask_question.py # Query NotebookLM167│ ├── notebook_manager.py # Library management168│ └── auth_manager.py # Google authentication169├── .venv/ # Isolated Python environment (auto-created)170└── data/ # Local notebook library171```172173When you mention NotebookLM or send a notebook URL, Claude:1741. Loads the skill instructions1752. Runs the appropriate Python script1763. Opens a browser, asks your question1774. Returns the answer directly to you1785. Uses that knowledge to help with your task179180---181182## Core Features183184### **Source-Grounded Responses**185NotebookLM significantly reduces hallucinations by answering exclusively from your uploaded documents. If information isn't available, it indicates uncertainty rather than inventing content.186187### **Direct Integration**188No copy-paste between browser and editor. Claude asks and receives answers programmatically.189190### **Smart Library Management**191Save NotebookLM links with tags and descriptions. Claude auto-selects the right notebook for your task.192193### **Automatic Authentication**194One-time Google login, then authentication persists across sessions.195196### **Self-Contained**197Everything runs in the skill folder with an isolated Python environment. No global installations.198199### **Human-Like Automation**200Uses realistic typing speeds and interaction patterns to avoid detection.201202---203204## Common Commands205206| What you say | What happens |207|--------------|--------------|208| *"Set up NotebookLM authentication"* | Opens Chrome for Google login |209| *"Add [link] to my NotebookLM library"* | Saves notebook with metadata |210| *"Show my NotebookLM notebooks"* | Lists all saved notebooks |211| *"Ask my API docs about [topic]"* | Queries the relevant notebook |212| *"Use the React notebook"* | Sets active notebook |213| *"Clear NotebookLM data"* | Fresh start (keeps library) |214215---216217## Real-World Examples218219### Example 1: Workshop Manual Query220221**User asks**: "Check my Suzuki GSR 600 workshop manual for brake fluid type, engine oil specs, and rear axle torque."222223**Claude automatically**:224- Authenticates with NotebookLM225- Asks comprehensive questions about each specification226- Follows up when prompted "Is that ALL you need to know?"227- Provides accurate specifications: DOT 4 brake fluid, SAE 10W-40 oil, 100 N·m rear axle torque228229230231### Example 2: Building Without Hallucinations232233**You**: "I need to build an n8n workflow for Gmail spam filtering. Use my n8n notebook."234235**Claude's internal process:**236```237→ Loads NotebookLM skill238→ Activates n8n notebook239→ Asks comprehensive questions with follow-ups240→ Synthesizes complete answer from multiple queries241```242243**Result**: Working workflow on first try, no debugging hallucinated APIs.244245---246247## Technical Details248249### Core Technology250- **Patchright**: Browser automation library (Playwright-based)251- **Python**: Implementation language for this skill252- **Stealth techniques**: Human-like typing and interaction patterns253254Note: The MCP server uses the same Patchright library but via TypeScript/npm ecosystem.255256### Dependencies257- **patchright==1.55.2**: Browser automation258- **python-dotenv==1.0.0**: Environment configuration259- Automatically installed in `.venv` on first use260261### Data Storage262263All data is stored locally within the skill directory:264265```266~/.claude/skills/notebooklm/data/267├── library.json - Your notebook library with metadata268├── auth_info.json - Authentication status info269└── browser_state/ - Browser cookies and session data270```271272**Important Security Note:**273- The `data/` directory contains sensitive authentication data and personal notebooks274- It's automatically excluded from git via `.gitignore`275- NEVER manually commit or share the contents of the `data/` directory276277### Session Model278279Unlike the MCP server, this skill uses a **stateless model**:280- Each question opens a fresh browser281- Asks the question, gets the answer282- Adds a follow-up prompt to encourage Claude to ask more questions283- Closes the browser immediately284285This means:286- No persistent chat context287- Each question is independent288- But your notebook library persists289- **Follow-up mechanism**: Each answer includes "Is that ALL you need to know?" to prompt Claude to ask comprehensive follow-ups290291For multi-step research, Claude automatically asks follow-up questions when needed.292293---294295## Limitations296297### Skill-Specific298- **Local Claude Code only** - Does not work in web UI (sandbox restrictions)299- **No session persistence** - Each question is independent300- **No follow-up context** - Can't reference "the previous answer"301302### NotebookLM303- **Rate limits** - Free tier has daily query limits304- **Manual upload** - You must upload docs to NotebookLM first305- **Share requirement** - Notebooks must be shared publicly306307---308309## FAQ310311**Why doesn't this work in the Claude web UI?**312The web UI runs skills in a sandbox without network access. Browser automation requires network access to reach NotebookLM.313314**How is this different from the MCP server?**315This is a simpler, Python-based implementation that runs directly as a Claude Skill. The MCP server is more feature-rich with persistent sessions and works with multiple tools (Codex, Cursor, etc.).316317**Can I use both this skill and the MCP server?**318Yes! They serve different purposes. Use the skill for quick Claude Code integration, use the MCP server for persistent sessions and multi-tool support.319320**What if Chrome crashes?**321Run: `"Clear NotebookLM browser data"` and try again.322323**Is my Google account secure?**324Chrome runs locally on your machine. Your credentials never leave your computer. Use a dedicated Google account if you're concerned.325326---327328## Troubleshooting329330### Skill not found331```bash332# Make sure it's in the right location333ls ~/.claude/skills/notebooklm/334# Should show: SKILL.md, scripts/, etc.335```336337### Authentication issues338Say: `"Reset NotebookLM authentication"`339340### Browser crashes341Say: `"Clear NotebookLM browser data"`342343### Dependencies issues344```bash345# Manual reinstall if needed346cd ~/.claude/skills/notebooklm347rm -rf .venv348python -m venv .venv349source .venv/bin/activate # or .venv\Scripts\activate on Windows350pip install -r requirements.txt351```352353---354355## Disclaimer356357This tool automates browser interactions with NotebookLM to make your workflow more efficient. However, a few friendly reminders:358359**About browser automation:**360While I've built in humanization features (realistic typing speeds, natural delays, mouse movements) to make the automation behave more naturally, I can't guarantee Google won't detect or flag automated usage. I recommend using a dedicated Google account for automation rather than your primary account—think of it like web scraping: probably fine, but better safe than sorry!361362**About CLI tools and AI agents:**363CLI tools like Claude Code, Codex, and similar AI-powered assistants are incredibly powerful, but they can make mistakes. Please use them with care and awareness:364- Always review changes before committing or deploying365- Test in safe environments first366- Keep backups of important work367- Remember: AI agents are assistants, not infallible oracles368369I built this tool for myself because I was tired of the copy-paste dance between NotebookLM and my editor. I'm sharing it in the hope it helps others too, but I can't take responsibility for any issues, data loss, or account problems that might occur. Use at your own discretion and judgment.370371That said, if you run into problems or have questions, feel free to open an issue on GitHub. I'm happy to help troubleshoot!372373---374375## Credits376377This skill is inspired by my [**NotebookLM MCP Server**](https://github.com/PleasePrompto/notebooklm-mcp) and provides an alternative implementation as a Claude Code Skill:378- Both use Patchright for browser automation (TypeScript for MCP, Python for Skill)379- Skill version runs directly in Claude Code without MCP protocol380- Stateless design optimized for skill architecture381382If you need:383- **Persistent sessions** → Use the [MCP Server](https://github.com/PleasePrompto/notebooklm-mcp)384- **Multiple tool support** (Codex, Cursor) → Use the [MCP Server](https://github.com/PleasePrompto/notebooklm-mcp)385- **Quick Claude Code integration** → Use this skill386387---388389## The Bottom Line390391**Without this skill**: NotebookLM in browser → Copy answer → Paste in Claude → Copy next question → Back to browser...392393**With this skill**: Claude researches directly → Gets answers instantly → Writes correct code394395Stop the copy-paste dance. Start getting accurate, grounded answers directly in Claude Code.396397```bash398# Get started in 30 seconds399cd ~/.claude/skills400git clone https://github.com/PleasePrompto/notebooklm-skill notebooklm401# Open Claude Code: "What are my skills?"402```403404---405406<div align="center">407408Built as a Claude Code Skill adaptation of my [NotebookLM MCP Server](https://github.com/PleasePrompto/notebooklm-mcp)409410For source-grounded, document-based research directly in Claude Code411412</div>413