Connect Gemini CLI to GTD Brain
If Gemini CLI is where you work, your task system should be one sentence away. Install the GTD Brain extension and capturing a thought — “add ‘fix the flaky auth test’ to my inbox” — happens without leaving the terminal, and “what can I do right now?” gets a real answer from your real board.
The extension points Gemini CLI at GTD Brain's hosted MCP (Model Context Protocol) server and adds a context file that teaches Gemini the GTD method plus /gtd: slash commands for capture, inbox zero, waiting-for, and the weekly review. Setup is one install command and a browser sign-in; your board stays the same board as the web, iOS, and Android apps, synced in real time.
What you’ll need
- A GTD Brain account — free to create at gtdbrain.com.
- Gemini CLI installed (the gemini command in your terminal), signed in to Google or with an API key.
Your connection address
The setup below uses this address — GTD Brain’s hosted MCP server:
Step-by-step setup
Install the GTD Brain extension
In a terminal, run the command below. Gemini CLI clones the extension from GitHub, shows the standard third-party notice, and enables it for every project.
gemini extensions install https://github.com/minosin/gtdbrain-gemini-extensionSign in with your email code
Start gemini and run the command below. A browser tab opens with GTD Brain's sign-in page: enter your email, type the code we send you, and click Allow. That's the whole handshake — no API key, no config file to edit. Over SSH, the CLI prints the sign-in URL to open by hand.
/mcp auth gtdbrainCheck the connection
Run /gtd:setup — it makes one free call to confirm GTD Brain answers. /mcp lists the gtdbrain server and its 16 tools; /extensions list shows the extension.
Try it
Ask Gemini “what's in my GTD Brain inbox?”, run /gtd:capture with something on your mind, or /gtd:what-now with where you are (calls, computer, errands, home). It reads and updates your real board through the same tools ChatGPT and Claude use.
What you can do once connected
The terminal connection shines for capture-without-context-switching and for turning work artifacts into tasks. Things people actually say mid-session:
- “/gtd:capture fix the flaky auth test”
- “Capture a card for each TODO comment you just found, with the file path in the note.”
- “/gtd:what-now computer”
- “I'm done for today — capture where I left off (reviewing the payment webhook) to my inbox.”
- “/gtd:waiting-for — anything I should chase before the weekend?”
- “/gtd:weekly-review”
Every change lands on the same board as the web, iOS, and Android apps — instantly. And whatever you ask, cards can be archived but never deleted, so nothing is ever lost to a misfired prompt.
Troubleshooting
gemini: command not found.
Install Gemini CLI first (npm install -g @google/gemini-cli, or see the Gemini CLI documentation), then rerun the install command.
/mcp shows gtdbrain as Disconnected or “requires authentication”.
Run /mcp auth gtdbrain to start the sign-in again. If no browser opens, copy the URL the CLI prints into any browser; the sign-in completes back in the terminal.
The /gtd: commands are missing.
Run gemini extensions list — the gtdbrain extension must be enabled. If a workspace has it disabled, gemini extensions enable gtdbrain turns it back on; gemini extensions update gtdbrain fetches the latest version.
Frequently asked questions
Can I add GTD Brain to Gemini CLI without the extension?
Yes. The extension is only configuration and prompt text — the MCP address above works with gemini mcp add --transport http gtdbrain <address> too. The extension adds the GTD context file and the /gtd: commands, and keeps itself up to date.
Is it safe to let an AI assistant touch my GTD board?
The connection is deliberately limited: your assistant can read, add, edit, move, and archive cards, but it can never delete one. Archived cards stay recoverable, so nothing is ever lost to a misfired prompt. You can disconnect at any time from the assistant's connector settings.
Do I need a paid GTD Brain subscription?
You can try it free: after connecting, your assistant can create your first 3 cards on a free GTD Brain account. Ongoing use of the connection needs a GTD Brain subscription — the same one subscription that covers the web, iOS, and Android apps.
What is MCP?
MCP (Model Context Protocol) is an open standard that lets AI assistants securely talk to outside tools and data. GTD Brain hosts an MCP server, so any MCP-capable assistant — ChatGPT, Claude, Claude Code, and others — can work with your board without a plugin or API key.
Do changes made by the assistant sync to my other devices?
Yes. The assistant works on the same board as the web, iOS, and Android apps, and every change syncs in real time. Add a card from a chat and it is on your phone before you switch apps.
Other connections
Connect ChatGPT · Connect Claude · Connect Claude Code · Connect Cursor · VS Code extension · Connect Telegram
New to GTD Brain? Get started free — create your board first, then connect Gemini CLI from this page in a couple of minutes.