·Minds Team

Connect Minds to Gemini CLI: extension and MCP setup

Install the Minds extension with gemini extensions install https://github.com/minds-ai-co/minds-mcp, or add https://getminds.ai/mcp with gemini mcp add, then sign in to Minds with OAuth. For Google Antigravity, follow the Antigravity guide.

Minds is the end-to-end platform for commercial synthetic research. With Gemini CLI you find your Audiences, plan a Study, review the proposed research plan and read results from your terminal. Gemini CLI talks to the Minds MCP server at https://getminds.ai/mcp, and you sign in with your Minds account over OAuth.

Requirements

  • Gemini CLI installed and signed in with a Gemini API key, Vertex AI or a Gemini Code Assist Standard or Enterprise license. Since June 2026, Gemini CLI no longer serves personal Google accounts; Google points individual users to Antigravity CLI, see Minds in Google Antigravity.
  • A Minds account. If you don't have one, you can sign up during the first connection.
  • A browser on the same computer for the OAuth sign-in, or a Minds API key for setups without a browser.

Which option fits?

OptionWhenSign-in
Minds extension (recommended)You want Minds plus its research instructions in one step. Listed in the Gemini CLI extensions gallery.OAuth with your Minds account
MCP serverYou manage MCP servers yourself, per project or for your user, or need an API key.OAuth or Minds API key
Google AntigravityYou use Antigravity 2.0, the Antigravity IDE or Antigravity CLI (agy).OAuth or Minds API key

Use one option per setup. The extension and a manual entry both register a server called minds, so keep only one of them.

Option A: install the Minds extension

1. Install

gemini extensions install https://github.com/minds-ai-co/minds-mcp

Gemini CLI shows what the extension adds: the remote MCP server minds at https://getminds.ai/mcp and the context file GEMINI.md, which tells the agent to clarify the research question and confirm a study plan before it starts a run. Confirm with Y. If the current folder is not trusted yet, Gemini CLI asks you to trust it first.

2. Check the installation

gemini extensions list

minds-mcp appears with the MCP server minds. Until you sign in, gemini mcp list shows the server as disconnected.

3. Sign in to Minds

Start gemini. Gemini CLI opens the Minds sign-in page in your browser the first time it connects; sign in and choose Allow. To start the sign-in yourself, or to sign in again later, run /mcp auth minds in the session. /mcp list then shows Minds as connected with its tools.

4. Keep it up to date

gemini extensions update minds-mcp

Option B: add Minds as an MCP server

Add the server for your user, so it is available in every folder:

gemini mcp add --scope user --transport http minds https://getminds.ai/mcp

Without --scope user, Gemini CLI writes the entry to .gemini/settings.json in the current project. You can also edit ~/.gemini/settings.json yourself. For streamable HTTP, use httpUrl; url selects the older SSE transport:

{
  "mcpServers": {
    "minds": { "httpUrl": "https://getminds.ai/mcp" }
  }
}

Start gemini and sign in as in step 3 of option A.

OAuth or API key?

  • OAuth (recommended): you sign in with your Minds account. Gemini CLI registers itself with Minds automatically, so you don't need a client ID or secret, and each person connects their own account.
  • API key: for setups without a browser, for example a CI job or a shared service account. Create a key in Settings → API Keys and add it as a header:
gemini mcp add --scope user --transport http --header "Authorization: Bearer minds_YOUR_API_KEY" minds https://getminds.ai/mcp

The key is stored in plain text in settings.json. Keep that file out of version control and prefer OAuth on personal machines.

Option C: Google Antigravity

Antigravity uses its own configuration: remote servers are set with serverUrl in ~/.gemini/config/mcp_config.json or .agents/mcp_config.json, and sign-in runs from Settings → Customizations → Authenticate. The Antigravity guide covers all three Antigravity surfaces.

Moving from Gemini CLI to Antigravity CLI? agy plugin import gemini converts the installed Minds extension into an Antigravity plugin with the same server.

Gemini app and Gemini Enterprise

Google also documents custom MCP connections in the Gemini app (personal accounts in the US) and in Gemini Enterprise. Minds can't be connected there yet. Use Gemini CLI or Google Antigravity until Minds supports these connectors.

How Minds works in Gemini CLI

  • Plan research first: before running anything, the agent presents the research plan for your approval. A Study starts only after you confirm.
  • Long runs continue on the server: Studies run durably on Minds, so a Study keeps running if you close the terminal. Ask Gemini CLI for its status later.
  • Go deeper in Minds: results link to the Study in Minds, where you review individual answers, compare Audiences and export results.

The tools Gemini CLI can use are listed in the tools reference. The MCP overview explains the integration as a whole.

Troubleshooting

Minds shows as disconnected. Run /mcp auth minds in a Gemini CLI session and complete the sign-in in the browser.

The sign-in page doesn't open. Gemini CLI receives the OAuth callback on a local port, so the browser has to run on the same computer. Over SSH or in a container, use an API key instead.

Two Minds servers or duplicate tools. The extension and a manual entry are both active. Remove one with gemini extensions uninstall minds-mcp or gemini mcp remove minds.

"Not authorized" for a Study or Audience you can open in Minds. Gemini CLI is connected to a different Minds account than the one that owns the Study. Run /mcp auth minds again with the right account, or share the Study with the connected account.

New Minds tools don't appear. Restart Gemini CLI so it reloads the tool list, and run gemini extensions update minds-mcp if you use the extension.

Other clients such as ChatGPT, Claude and Codex are listed in the integrations overview.

MCP server URL

https://getminds.ai/mcp

View full MCP documentation