NotebookLM AI Plugin

Drive Google NotebookLM (https://notebook.google.com) from the terminal. Both notebook.google.com and notebooklm.google.com notebook URLs are accepted.

Supports:

• Chat with Notebook AI (source-grounded Q&A with citations)
• Slide Deck generation (PPTX/PDF)
• Audio Overview (M4A — deep dive, brief, critique, debate formats)
• Video Overview (MP4 — auto, classic, whiteboard, kawaii, anime, watercolor, retro_print, heritage, paper_craft styles)
• Mind Map (JSON tree)
• Flashcards / Quiz (interactive HTML)
• Infographic (PNG — landscape, portrait, square; 11 visual styles)
• Report (Markdown — briefing doc, study guide, blog post, custom prompt)
• Data Table (CSV)
• Source management (add URLs, YouTube, files, pasted text; list, delete)
• Fast/Deep web research with auto-import
• Notes management (create, update, delete, list)
• Notebook library management (add, list, search, activate) + listing your account's notebooks

All commands run from this skill directory:

cd <skill-dir>   # the folder containing SKILL.md
npx -y bun scripts/main.ts <command> [options]

Quick start

# 1. Authenticate (opens Chrome for Google login; cookies are cached)
npx -y bun scripts/main.ts login

# 2. Check the session and see your notebooks (ids + source counts)
npx -y bun scripts/main.ts notebooks remote

# 3. Save one as the default
npx -y bun scripts/main.ts notebooks add https://notebook.google.com/notebook/<id> --name "My Research"

# 4. Use it
npx -y bun scripts/main.ts chat --question "What are the key findings?"
npx -y bun scripts/main.ts generate slide_deck --output slides.pdf
npx -y bun scripts/main.ts generate audio --format deep_dive --length long
npx -y bun scripts/main.ts generate quiz --difficulty medium --quantity more --output quiz.html
npx -y bun scripts/main.ts generate data_table --instructions "Compare the frameworks" --output table.csv

Authentication

login launches Chrome (with a dedicated profile under the data dir) and waits for you to sign in at notebook.google.com, then captures the Google session cookies. Cookies expire after a few weeks; any command that hits a login redirect tells you to run login --force.

npx -y bun scripts/main.ts login            # reuse cached cookies if still valid
npx -y bun scripts/main.ts login --force    # always re-open Chrome
npx -y bun scripts/main.ts notebooks remote # fast auth check: lists notebooks from your account

Every RPC call fetches a fresh CSRF token (SNlM0e) and session id (FdrFJe) from the app page, so nothing else needs refreshing. If the origin changes again, set NOTEBOOKLM_BASE_URL (or edit scripts/constants.ts).

Commands

Notebook management

npx -y bun scripts/main.ts notebooks remote                 # notebooks on your Google account (live)
npx -y bun scripts/main.ts notebooks list                   # local library
npx -y bun scripts/main.ts notebooks add <url> [--name <name>] [--description <desc>] [--topics <t1,t2>]
npx -y bun scripts/main.ts notebooks remove <id>
npx -y bun scripts/main.ts notebooks activate <id>          # default for every other command
npx -y bun scripts/main.ts notebooks search <query>

Chat with Notebook AI

npx -y bun scripts/main.ts chat --question "What are the key findings?" --notebook <url|id>
npx -y bun scripts/main.ts chat --question "Tell me more about that" --conversation-id <id>
npx -y bun scripts/main.ts chat --question "Summarize the methodology" --json   # answer + citations

Chat uses all ready sources in the notebook. Citations reference source ids (see sources list).

Source management

npx -y bun scripts/main.ts sources list --notebook <url|id>          # id, title, type, status, url
npx -y bun scripts/main.ts sources add-url https://example.com/article
npx -y bun scripts/main.ts sources add-youtube https://youtube.com/watch?v=xxx
npx -y bun scripts/main.ts sources add-text --title "My Notes" --content "Important findings..."
npx -y bun scripts/main.ts sources add-file ./paper.pdf
npx -y bun scripts/main.ts sources delete <sourceId>

New sources show status: processing for a little while; generation and chat only use sources whose status is ready. Supported files: PDF, TXT, MD, DOCX, PPTX, CSV, EPUB, images (PNG, JPG, WEBP), audio/video.

Research

npx -y bun scripts/main.ts research fast --query "latest AI agent frameworks"   # ~1–3 min
npx -y bun scripts/main.ts research deep --query "state of LLM reasoning"       # ~5–10 min, produces a report
npx -y bun scripts/main.ts research fast --query "topic" --import                # import found sources
npx -y bun scripts/main.ts research status

Notes

npx -y bun scripts/main.ts notes list
npx -y bun scripts/main.ts notes create --title "Key Takeaways" --content "1. Finding one..."
npx -y bun scripts/main.ts notes update <noteId> --title "Updated" --content "New content"
npx -y bun scripts/main.ts notes delete <noteId>

Artifacts

npx -y bun scripts/main.ts artifacts list                 # everything in the Studio panel, with status
npx -y bun scripts/main.ts generate <type> [options]

Types: slide_deck, audio, video, mind_map, flashcards, quiz, infographic, report, data_table

Generation is asynchronous on Google's side: the command creates the artifact, polls the Studio list until it is completed and its download URL is available (audio/video can take 10–20 minutes), then downloads it. If auto-download fails, the URL is printed so it can be opened in a browser.

Options

Global options

Per-type options

Output formats

Environment variables

Rate limits (free tier)

Quota rejections surface as an RPCError of kind rate_limit.

Troubleshooting

Tests

cd test
npx -y bun install                          # dev deps only (typescript, types)
npx -y bun x tsc --noEmit                   # type-check
npx -y bun test unit                        # offline: payload shapes + decoders against recorded responses
npx -y bun test live.test.ts                # live: auth + read endpoints (skips without a session)
NOTEBOOKLM_LIVE_WRITE=1 npx -y bun test live.test.ts             # + add/delete a text source and a note
NOTEBOOKLM_LIVE_GENERATE=quiz npx -y bun test live.test.ts       # + generate one artifact end-to-end

Skill-level evals (prompts + assertions for the skill-creator harness) live in test/evals.json.