Kortex MCP vs notebooklm-mcp-cli for Gemini Notebook (formerly NotebookLM)
Kortex MCP vs notebooklm-mcp-cli for connecting Claude, ChatGPT or Cursor to NotebookLM: sign-in, clients, tools, risks and price. Checked Nov 4, 2026.
On this page
Short answer: Kortex MCP vs notebooklm-mcp-cli is hosted versus self-run. notebooklm-mcp-cli is a free, MIT-licensed command-line tool and MCP server that signs in to Gemini Notebook (formerly NotebookLM) with cookies taken from your browser. Kortex MCP comes with Kortex Pro and runs each call through the Kortex extension in your own Chrome.
The Model Context Protocol (MCP) is the open standard that AI apps like Claude, ChatGPT and Cursor use to call outside tools. notebooklm-mcp-cli is an open-source Python project that gives those apps Gemini Notebook tools, plus a command line. Kortex MCP is a hosted MCP server that comes with Kortex Pro and works through the Kortex Chrome extension. We built Kortex MCP and publish this blog. Every fact about notebooklm-mcp-cli below links to its GitHub README, its own guides or PyPI, and we checked each one on November 4, 2026. Both connect to the same notebooks in your Google account. So the choice is about setup, safety and cost, not content. Below: who should pick which, a 14-point table, where each one wins, and prices side by side.
Which should you pick: Kortex MCP or notebooklm-mcp-cli?
Pick notebooklm-mcp-cli if you want free, open source and a terminal; pick Kortex MCP if you want Claude or ChatGPT connected through a sign-in page, with no cookies copied. notebooklm-mcp-cli is MIT-licensed and free (GitHub, Nov 4, 2026). Kortex MCP is included with Kortex Pro, from $108 a year.
notebooklm-mcp-cli fits you if:
- You want free and open source, and you’re at home in a terminal.
- You want a command-line tool for scripts and servers, and an MCP server that runs without a browser window open.
- You’re on a Gemini Notebook Enterprise account, which it supports experimentally.
Kortex MCP fits you if:
- You want Claude (web or desktop) or ChatGPT connected through a sign-in page, with no config files.
- You’d rather no Google cookies are copied anywhere. Calls run in the Chrome you’re already signed in to.
- You want the Kortex extension too: imports, organizing, exports and automations.
Kortex MCP vs notebooklm-mcp-cli: how do they compare?
The two take opposite routes to the same notebooks. notebooklm-mcp-cli calls Gemini Notebook’s internal APIs from your computer with saved cookies; Kortex MCP sends each call to the Kortex extension in your Chrome. The table compares 14 points. Every Kortex MCP feature needs Kortex Pro or Lifetime.
| Feature | Kortex MCP | notebooklm-mcp-cli |
|---|---|---|
| What it is | A hosted MCP endpoint that relays each call to the Kortex extension in your Chrome | A Python command-line tool (nlm) and an MCP server you install and run |
| Price | Included with Kortex Pro and Lifetime | Free, MIT license |
| Install | The Kortex Chrome extension; nothing else | uv tool install notebooklm-mcp-cli, with Python 3.11 or later |
| Google sign-in | Uses the session already in your Chrome; Kortex never stores your Google password or cookies | Opens your browser to log in, then extracts and saves your cookies; they last about 2 to 4 weeks and refresh through a headless browser if a profile is saved |
| How it reaches Gemini Notebook | Through the Kortex extension, signed in as you | Undocumented internal APIs, which its README says may change without notice |
| Connecting a client | Paste one URL and sign in with OAuth, or use an API key | A setup wizard (nlm setup) writes the config for tools it finds on your computer |
| Claude on the web | Custom connector, on the web and in the desktop app | Its remote guide says it is “not currently a secure, turnkey remote MCP service” |
| ChatGPT | MCP app in developer mode | The ChatGPT desktop app, through the wizard |
| Coding tools | Claude Code, Cursor, VS Code, Windsurf, Cline, Codex CLI, Antigravity, Zed | Claude Code, Cursor, Windsurf, Cline, Codex CLI, Antigravity, Gemini CLI, GitHub Copilot in VS Code, OpenCode |
| Tools | 145+, including Kortex’s tags, collections, podcasts and schedules | 55 MCP tools, plus the nlm command line |
| What must be running | Chrome, with the extension signed in; minimised is fine. The first call after idle can take up to 30 seconds | Nothing after sign-in; a saved browser profile handles automatic re-login |
| Access controls | Read-only API keys by default, with a profile per key | Not listed in its README; its remote guide says its HTTP endpoint has no built-in authentication |
| Scheduled runs | Hourly, daily or weekly notebook jobs while Chrome is open | Multi-step pipelines you start from the CLI or MCP |
| Gemini Notebook Enterprise | Not covered in the Kortex MCP docs | Experimental support |
notebooklm-mcp-cli facts come from its GitHub README, remote MCP guide, MCP guide and PyPI page, checked on November 4, 2026. Kortex MCP facts come from the Kortex MCP docs.
Where does notebooklm-mcp-cli win?
notebooklm-mcp-cli wins on cost, openness and running unattended. It’s free under the MIT license, it shipped version 0.16.0 on November 4, 2026 (PyPI), and after one sign-in it runs with no browser window open. Kortex MCP needs a paid plan and a running Chrome.
Free and open source. It’s MIT-licensed, with 6,457 stars on GitHub and commits as recent as November 4, 2026. The author takes support through Buy Me a Coffee. Kortex MCP needs Kortex Pro or Lifetime.
A real command line. nlm handles notebooks, sources, Studio items, pipelines and downloads from a terminal, so it fits scripts and servers. Kortex MCP takes an API key for scripts, but it has no command-line tool of its own.
No browser window to keep open. After you sign in once, it runs headless and refreshes its login through a headless browser when a profile is saved. When Google’s login fully expires, you run nlm login again. Kortex MCP needs Chrome running with the extension signed in.
More client setups, and Enterprise. Its wizard also configures Gemini CLI, GitHub Copilot and OpenCode, and its README lists experimental support for Gemini Notebook Enterprise. The Kortex MCP docs don’t cover Enterprise accounts.
Where does Kortex MCP win?
Kortex MCP wins on sign-in safety and web clients; we built it so nothing about your Google account leaves your browser. Kortex never stores your Google password or cookies, and Claude on the web connects with one URL and a sign-in. notebooklm-mcp-cli’s remote guide says it is “not currently a secure, turnkey remote MCP service” (GitHub, Nov 4, 2026).
No cookies to copy. Calls run through the Kortex extension in the Chrome you already use, and nothing about your Google account leaves your browser. notebooklm-mcp-cli saves your Google cookies to a file, or encrypted in your system keychain if you switch that on.
Claude and ChatGPT on the web. Add one URL as a custom connector in Claude and sign in, or add it as an MCP app in ChatGPT’s developer mode. Our guide to connecting Gemini Notebook to Claude and ChatGPT walks through both.
Safer defaults. New Kortex API keys are read-only until you choose otherwise, and each key has its own profile. notebooklm-mcp-cli’s README lists no access controls, and its remote guide says its HTTP endpoint has no built-in authentication.
More than notebooks. Its 145+ tools also reach Kortex’s own layer, such as tags, collections, podcast feeds, templates and scheduled jobs. Some answer with a panel you use right in the chat. Schedules can run a notebook job hourly, daily or weekly; here’s a daily briefing recipe.
Both tools serve a builder-heavy crowd. When we analyzed the Kortex MCP waitlist survey (776 respondents, Sep-Oct 2026), Claude was the assistant named most, 126 times, ahead of ChatGPT at 69, and 67 people named a coding or agent tool.
Hand your notebooks to your assistant, no cookies copied
Kortex MCP adds to Claude, ChatGPT, Cursor or VS Code as one URL and a sign-in, with 145+ notebook tools. It's included with Kortex Pro and Lifetime.
How much do Kortex MCP and notebooklm-mcp-cli cost?
notebooklm-mcp-cli is free, with tips welcome through Buy Me a Coffee (README, Nov 4, 2026). Kortex MCP is included with Kortex Pro, at $9.99 a month or $108 a year, and with Lifetime at $199 once. The Kortex extension itself is free. So what does the subscription pay for? Kortex runs the hosted endpoint and the OAuth sign-in, and Pro also unlocks the rest of the extension: unlimited imports and exports, bulk import, automations, Google Docs sync, sync across devices and a podcast feed with 1,000 minutes. notebooklm-mcp-cli costs nothing, but you install, update and sign it in yourself, and its README asks you to accept the risks of undocumented APIs. Every Kortex plan has a 7-day money-back guarantee, so you can try Kortex MCP and ask for a refund within that window if it doesn’t fit.
| Kortex MCP | notebooklm-mcp-cli | |
|---|---|---|
| Price | Included with Kortex Pro: $9.99 a month or $108 a year, or $199 Lifetime | Free; the author takes tips through Buy Me a Coffee |
| Without paying | The Kortex extension is free; Kortex MCP needs Pro | Everything |
| Money-back guarantee | 7 days | Not applicable |
Kortex prices are on the pricing section; many countries see a lower Kortex price at checkout.
Can you try both?
Yes. Both reach the same notebooks in your Google account, so you can try one without moving anything. While you compare, connect one at a time so you always know which server answered. Neither changes how your notebooks are stored, so removing either one later leaves your notebooks exactly as they were. For example, connect notebooklm-mcp-cli to Claude Code through its setup wizard and ask a question about one notebook, then remove it, add Kortex MCP as a custom connector and ask the same question. Which setup fits your day better? That’s the test that matters. The wizard backs up each config file it changes, so undoing its setup is straightforward. For the wider field, including archived and browser-automation servers, read our Gemini Notebook MCP server guide.
One URL, a sign-in, and 145+ notebook tools
Kortex MCP connects Claude, ChatGPT, Cursor and Codex to Gemini Notebook through your own Chrome. It's included with Kortex Pro and Lifetime.
Frequently asked questions
Is there an official Gemini Notebook MCP server?
No. Google ships none, and the request for an official NotebookLM MCP server, issue 19 on the google/mcp GitHub repository, is still open as of November 4, 2026. The options are community projects like notebooklm-mcp-cli and services like Kortex MCP. Our Gemini Notebook MCP guide covers them all.
How do I install notebooklm-mcp-cli?
Its README installs it with uv tool install notebooklm-mcp-cli, and PyPI lists Python 3.11 or later. Signing in opens your browser so the tool can extract your Google cookies, and nlm setup then writes the config for the AI tools it finds on your computer, backing up each file it changes.
What does notebooklm-mcp-cli say about its risks?
Its README says the internal APIs it uses “are undocumented and may change without notice”, that it “may not be aligned with Google’s Terms of Service”, and to use it “at your own risk for personal/experimental purposes”. It offers optional encrypted login storage in your system keychain.
Can I use Kortex MCP in Claude.ai or ChatGPT?
Yes. Kortex MCP is a remote endpoint with OAuth sign-in, so it adds to Claude on the web and desktop as a custom connector, and to ChatGPT as an MCP app in developer mode. It’s included with Kortex Pro and Lifetime, and needs Chrome running with the Kortex extension signed in.
Where did notebooklm-mcp-cli go on GitHub?
The repository now lives at jacob-bd/gemini-notebook-mcp-cli, and the old jacob-bd/notebooklm-mcp-cli address redirects there. The package you install from PyPI is still called notebooklm-mcp-cli, and the command-line tool is still nlm.
What about PleasePrompto/notebooklm-mcp?
It was the best-known browser-automation server for Gemini Notebook, with 3,443 GitHub stars, but its repository is archived, so it gets no updates or fixes. notebooklm-mcp-cli and Kortex MCP are both maintained alternatives, as of November 4, 2026.
Editorial note: fact-checked against the official help pages and docs linked above on November 4, 2026. Spotted something out of date? Contact us and we will fix it, or read more about Kortex.
