notebooklm-py

notebooklm-py

A Comprehensive Google Gemini Notebook Skill & Unofficial Python API. Full programmatic access to NotebookLM's features—including capabilities the web UI doesn't expose—via Python, CLI, and AI agents like Claude Code, Codex, and OpenClaw.

Note (July 2026): Google rebranded NotebookLM to Gemini Notebook. It remains the same standalone product (now also reachable inside the Gemini app), existing links redirect automatically, and this library drives the same underlying service and works unchanged. The package keeps the notebooklm-py name.

Source & Development: https://github.com/teng-lin/notebooklm-py

⚠️ Unofficial Library - Use at Your Own Risk

This library uses undocumented Google APIs that can change without notice.

• Not affiliated with Google - This is a community project

• APIs may break - Google can change internal endpoints anytime

• Rate limits apply - Heavy usage may be throttled

Best for prototypes, research, and personal projects. See Troubleshooting for debugging tips.

What You Can Build

🤖 AI Agent Tools - Integrate NotebookLM into Claude Code, Codex, and other LLM agents. Ships with a root NotebookLM skill for GitHub and npx skills add discovery, local notebooklm skill install support for Claude Code and .agents skill directories, and repo-level Codex guidance in AGENTS.md.

📚 Research Automation - Bulk-import sources (URLs, PDFs, YouTube, Google Drive), run web/Drive research queries with auto-import, and extract insights programmatically. Build repeatable research pipelines.

🎙️ Content Generation - Generate Audio Overviews (podcasts), videos, slide decks, quizzes, flashcards, infographics, data tables, mind maps, and study guides. Full control over formats, styles, and output.

📥 Downloads & Export - Download all generated artifacts locally (MP3, MP4, PDF, PNG, CSV, JSON, Markdown). Export to Google Docs/Sheets. Features the web UI doesn't offer: batch downloads, quiz/flashcard export in multiple formats, mind map JSON extraction.

Use Cases & Recipes

NotebookLM is a grounded engine: Gemini does the heavy reading and answers from your sources with citations. The winning pattern is to let it do the expensive analysis while your agent (Claude Code, Codex, …) orchestrates and handles the final mile — using NotebookLM as a zero-token synthesis + memory layer an agent drives in a loop, and pulling structured artifacts out in bulk and in richer, scriptable formats. Recipes people build on top of this library, grouped by what they use NotebookLM as:

Spend fewer tokens — let NotebookLM do the expensive thinking:

• 🪙 Zero-token research offload — Throw 30 documents into a notebook, let Gemini do the heavy analysis, and have your agent spend tokens only on the final polish. The agent just orchestrates (create → source add → ask); the reasoning happens server-side. In the wild: a four-workflow guide to stop Claude Code burning tokens on NotebookLM.

• 🧠 Knowledge distillation → a permanent skill — Run Deep Research (source add-research "your topic" --mode deep) or load a doc corpus, let NotebookLM's Gemini condense it, and bake the result into a SKILL.md your agent loads at startup — build once, reuse with zero runtime tokens or network calls, git-versioned and immune to UI drift. A packaged domain expert without hand-curating sources. (Dumping raw docs into a skill flattens the hierarchy; NotebookLM condensing first is what makes it work.)

• ✅ Self-validating skills — Have NotebookLM generate the eval set — a quiz straight from your sources — to grade an agent skill against ground truth instead of test questions you'd bias yourself. Build the skill, run it against the NotebookLM-authored evals, iterate to a pass. In the wild: a skill that scored 4/10 on the first pass and 10/10 after one iteration, graded by a NotebookLM-generated quiz.

Give your agent memory — persistent, grounded recall:

• 💾 Persistent cross-session memory — Keep a "Master Brain" notebook; a wrap-up step appends each session's decisions and fixes as notes (note create / ask --save-as-note), and a line in your CLAUDE.md queries it (ask) at the start of the next session. Storage and recall live on Google's infrastructure.

• 🧩 Grounded memory for coding agents — Expose a notebook of your internal docs/RFCs/architecture over the MCP server (or plain ask) so an agent answers from your code with citations rather than plausible-sounding guesses — a zero-infra alternative to standing up your own vector DB and embedding pipeline. In the wild: turning a notebook into the source-grounded "project brain" a coding agent consults before it writes code.

• 🪞 Query your own notes / journal — Load years of daily notes, meeting logs, or a journal and ask for cited answers across your own history — surfacing long-term patterns a keyword search can't (e.g. a weekly summary synthesized from 282 daily notes, every claim linked back to the entry it came from). In the wild: chatting with a year of daily notes as a cited knowledge base.

Turn your sources into answers & artifacts — cited responses, generated media, and exports:

• 📞 Grounded knowledge base / troubleshooting oracle (RAG) — Load product docs, FAQs, RFCs, and past tickets, then ask --json for source-grounded, cited answers for support, on-call, or internal Q&A. Or have an agent point it at an entire fast-moving tool's docs — more than the agent can hold in context — as a troubleshooting oracle it queries the moment it hits an error. In the wild: OpenClaw drove the library to scrape all 524 pages of docs.openclaw.ai, dedupe the duplicate translations, and audit it down to 269 clean sources (missing/extra/duplicate = 0).

• 🔁 Multi-format content repurposing — One source set, every format: generate audio (podcast), generate video, generate slide-deck, plus a generate report blog draft, generate quiz, and generate flashcards — fan a single notebook out across channels.

• 📤 Bulk, scriptable exports — Pull mind maps as JSON, flashcards/quizzes as JSON/Markdown/HTML, data tables as CSV, and reports as Markdown — in bulk, to local files, straight into Anki, your mind-mapping tool, or a repo (download / download --all). The programmatic "get data out" half of the library, not just "put sources in."

• 🕸️ Obsidian / knowledge-graph sync — Run the CLI from your vault root so downloaded artifacts (reports, mind-map JSON, transcripts) land as files in your knowledge graph; community skills built on this library even resolve NotebookLM's citation markers into Obsidian [[wikilinks]]. Pair with a podcast overview for an audio digest of your notes. In the wild: "Claude Code + NotebookLM + Obsidian = GOD MODE".

Run it unattended, at scale, or on the go — scheduled, headless, and remote:

• 🚨 Incident runbook generator — On an alert, spin up a notebook of the relevant docs, ask targeted diagnostic questions, and generate a briefing-doc report (generate report --format briefing-doc --wait, then download report) as an automated runbook.

• 📚 Curriculum / study-set builder — Scrape a syllabus or developer roadmap, create one notebook per topic (with deliberate pacing to dodge rate limits), and bulk-generate podcasts, quizzes, and flashcards for each.

• 📰 Scheduled audio briefings — Pair auth refresh --quiet (cron/launchd/systemd) with generate audio to publish a fresh personalized briefing to a podcast feed on a schedule.

• 📱 NotebookLM from your phone, agent-driven — Self-host the remote MCP connector behind a Cloudflare/Tailscale tunnel and add it as a custom connector on the web (claude.ai Connectors, or ChatGPT with Developer Mode). Then drive the full toolset — deep research, source ingestion, studio generation, cited Q&A — from the claude.ai mobile app on the go (ChatGPT's MCP connectors are web-only), chained with your other MCP tools, instead of app-hopping.