NotebookLM skill
Research with NotebookLM from Claude Code: cited answers, podcasts, slides and quizzes.
Community, verifiedby Claude WorldChecked October 27, 2026
Install NotebookLM free
Leave your email and the install steps open right here: one prompt to paste into your AI tool. Works for all 151 skills.
A community NotebookLM skill for Claude Code and other coding agents, by Claude World. It builds notebooks from your links and files, asks NotebookLM questions with citations and turns the sources into podcasts, slides or study material. We list it instead of PleasePrompto's older NotebookLM skill, whose repository was archived in September 2026.
Who it is for
- Researchers
- Content creators
- Students
- Marketers
What's inside NotebookLM
This skill is the guide to a full Python package: its 231-line SKILL.md teaches Claude the commands of notebooklm-skill, which ships a command-line tool, five pipelines and a 13-tool MCP server in the same repo. The package reaches NotebookLM through notebooklm-py, an unofficial library that uses your browser login instead of a Google API key.
When it switches on
It switches on when you ask for NotebookLM by name, for analysis with cited sources, for research that turns into content, or for podcasts, slides, study material, RSS digests or trend research.
What the instructions cover
- Signing in once through the browser with notebooklm-auth, named profiles to keep accounts apart, and a rule never to open or share the saved session file
- Notebook commands: create from mixed sources, list, add one source, summarize and ask, preferring notebook IDs over titles in repeated runs
- Eleven output types, from audio and cinematic video to slides, quizzes, flashcards, mind maps and data tables, with jobs that can run in the background and download later
- Five pipelines that turn sources into an article draft, social drafts, an RSS digest with Q&A, trend-based content or a full set of outputs; none of them publishes anything
- Eight operating rules, such as checking that sources really loaded, reporting partial failures and never deleting or overwriting without your say-so
What it needs
- The notebooklm-skill package installed with pip, uvx or the repo's install.sh, which also links five commands into ~/.local/bin
- Python 3.10 or newer and Chromium for the first login, or your own Chrome or Edge
- A Google account with NotebookLM access; no API key
- Optional: the trend-pulse command for trend research, and ffmpeg plus Poppler to turn slides and a podcast into a video
| File | What it is |
|---|---|
SKILL.md | 231 lines: sign-in, the core commands, output types, pipelines, MCP tools and operating rules |
scripts/ | 7 files (2,349 lines): the command-line tool, pipelines, sign-in helper, skill installer and a script that turns slides and a podcast into a video |
mcp_server/ | 3 files (689 lines): the MCP server and its 13 tools |
references/ | 3 guides (394 lines): the unofficial API it relies on, the JSON output format and pipeline recipes |
tests/ | 11 files (1,792 lines) of offline tests, including checks against the notebooklm-py interface |
install.sh | 85 lines: a separate Python environment, Chromium, five commands in ~/.local/bin and the skill in ~/.claude/skills |
README.md | 221 lines, with a Traditional Chinese copy; docs/ has setup guides in both languages and examples/ three walk-throughs in Chinese |
CHANGELOG.md | 195 lines of release notes back to 1.0.0; also SECURITY.md, AGENTS.md, LICENSE (MIT), pyproject.toml and GitHub workflows for tests and releases |
Watch out: It works through notebooklm-py, an unofficial library built on NotebookLM's private web API, so Google can change or block it without notice, and quotas and generation times vary. The saved browser session works like your Google password: keep it private, and consider a separate Google account for automation.
Prompts to try
Make a NotebookLM notebook from these three articles and tell me where they disagreeTurn my research notebook into a podcast episode and save it in my output folderCreate flashcards and a quiz from the PDF of chapter fourHow to install it
Skills follow the open Agent Skills standard, so this one is not tied to one company. Pick the AI tool you use. Not sure? The first tab covers most of them in one step.
Install this skill for me so it works in every project: run `npx skills add https://github.com/claude-world/notebooklm-skill -g -y` in my terminal. If Node.js is missing, tell me how to install it first. When it's done, confirm the skill is installed and tell me how to start using it.Or do it yourself:
One command installs the skill: it finds the AI tools on your computer and asks which ones should get it.
npx skills add https://github.com/claude-world/notebooklm-skill -gYou need Node.js installed (free, from nodejs.org). The -g flag makes the skill available in every project.
Good to know: The skill is a guide to the notebooklm-skill command-line tools, so install those too: run ./install.sh from a clone of the repo (it also adds the skill to Claude Code) or pip install notebooklm-skill followed by python -m playwright install chromium, then sign in to Google with notebooklm-auth setup.
Unlock the install steps
Free. Leave your email once to open the install steps for all 151 skills, and get the best new skills each month.
How we checked it
On October 27, 2026 we read all 54 files in the skill's folder, starting with its 231-line SKILL.md. The section on what's inside comes from those files. Its license is MIT. Its files were last changed in October 2026. We have not run it on a test task yet.
Want one we ran on a real task? See the skills we tested in education and learning, ranked by score.