By hah23255
Conversational tool-calling for NotebookLM via the notebooklm-mcp-cli MCP server (~35 tools: notebooks, sources, studio artifacts, research, downloads, sharing, batch, cross-notebook). Heavier than the skills plugin — its tools load into every session's context. Pinned to notebooklm-mcp-cli 0.7.3 (first post-SSRF-revert release). Planned migration: the official notebooklm-py MCP server (25 tools) once released. Requires Python 3.11+ and uv; authenticate with 'nlm login' after install.
Based on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Google NotebookLM automation for Claude Code / Claude Desktop / Cowork — skills, plugin, and optional MCP server in one marketplace.
Built on teng-lin/notebooklm-py (skills + CLI) and jacob-bd/notebooklm-mcp-cli (MCP server). Unofficial integrations — not endorsed by Google.
| Pillar | Where | How it activates |
|---|---|---|
| Skill | .claude/skills/notebooklm/ + .claude/skills/run-notebooklm/ | Clone this repo → /notebooklm and /run-notebooklm work immediately in Claude Code |
| MCP | .mcp.json (project-scope) and the notebooklm-mcp plugin (user-scope) | Clone → Claude Code offers the project MCP server; or install the plugin on any machine |
| Plugin | plugins/ + .claude-plugin/marketplace.json | /plugin marketplace add + /plugin install — no clone needed |
| Plugin | Contents | Context cost | Install when… |
|---|---|---|---|
notebooklm | /notebooklm skill (full API: notebooks, sources, chat, podcasts, videos, slide decks, infographics, quizzes, flashcards, mind maps) · /run-notebooklm verify/troubleshoot runbook with smoke-test driver · SessionStart hook that auto-installs the CLI | ~165 always-on tokens | Always — this is the core |
notebooklm-mcp | MCP server wiring (≈35 tools via notebooklm-mcp-cli, pinned 0.7.3) | 35 tool schemas in every session | Only if you want conversational tool-calling instead of / alongside the skill |
Why two plugins in one repo: they share one version train and one sync script,
but the MCP server's tool schemas cost context in every session — so it's
opt-in rather than bundled. The root .mcp.json gives the same server to
anyone who clones the repo (project scope), without installing any plugin.
Verification status: skills, hook, sync script, marketplace, and both plugin installs are verified end-to-end in a clean container. The MCP server launch (
uvx --from notebooklm-mcp-cli==0.7.3 notebooklm-mcp) is configuration-verified (Claude Code accepts and attempts it); a full first start needs a machine allowed to fetch the PyPI package, plusuv(install) andnlm loginfor auth.
/plugin marketplace add hah23255/notebooklm-claude-toolkit
/plugin install notebooklm@notebooklm-toolkit
/plugin install notebooklm-mcp@notebooklm-toolkit # optional
Non-interactive: claude plugin marketplace add hah23255/notebooklm-claude-toolkit && claude plugin install notebooklm@notebooklm-toolkit
pip install "notebooklm-py[browser]" # pip3 / python3 -m pip on Debian/Ubuntu
playwright install chromium # ~130 MB; needed for login, not for env-var auth
notebooklm login # opens a browser for Google sign-in
notebooklm doctor # verify: all checks pass
notebooklm login --browser msedgeexport NOTEBOOKLM_AUTH_JSON="$(cat ~/.notebooklm/profiles/default/storage_state.json)"nlm login (separate credential store).storage_state.json like a password. Keep it out of synced folders
(OneDrive/Dropbox); the default location ~/.notebooklm/ is fine.Verify the whole stack any time: /run-notebooklm (or
bash .claude/skills/run-notebooklm/smoke.sh — exit 0 ready · 2 unauthenticated
· 3 broken install).
.claude/skills/ ← CANONICAL skills (edit here)
│ ├── notebooklm/ ← vendored from notebooklm-py (notebooklm skill install)
│ └── run-notebooklm/ ← verify/troubleshoot runbook + smoke.sh driver
.claude-plugin/marketplace.json
plugins/
│ ├── notebooklm/ ← skills plugin (mirrors of canonical + hook)
│ │ ├── .claude-plugin/plugin.json
│ │ ├── hooks/hooks.json ← SessionStart → ensure-notebooklm.sh
│ │ ├── scripts/ensure-notebooklm.sh ← pip→pip3→python3 -m pip fallback
│ │ ├── scripts/sync-skills.sh ← canonical → plugin sync + version check
│ │ └── skills/
│ └── notebooklm-mcp/ ← MCP plugin (config only; server code stays upstream)
│ └── .claude-plugin/plugin.json ← uvx --from notebooklm-mcp-cli==0.7.3
docs/ ← ecosystem evaluation + modernization plan
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimnpx claudepluginhub hah23255/notebooklm-claude-toolkit --plugin notebooklm-mcpFull programmatic access to Google NotebookLM: create notebooks, add sources (URLs, YouTube, PDFs, Drive), chat with citations, and generate/download audio overviews, videos, slide decks, infographics, quizzes, flashcards, and mind maps. Bundles the notebooklm-py agent skill, a run/verify skill, and a SessionStart hook that installs the CLI. Version tracks the vendored notebooklm-py skill.
Harness-native ECC plugin for engineering teams - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Browser automation and end-to-end testing MCP server by Microsoft. Enables Claude to interact with web pages, take screenshots, fill forms, click elements, and perform automated browser testing workflows.
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.