Quick Start

Two things to install: Google's Antigravity CLI (agy), and this server, which drives it. Then one check that tells you both worked.

Node 18.19 or newer.

1. Install the CLI and sign in

curl -fsSL https://antigravity.google/cli/install.sh | bash
agy

On Windows, use the official installer from https://goo.gle/gemini-cli-migration instead of the curl line.

Running agy on its own walks you through signing in. That happens once, and it is interactive. The relay never signs in for you.

2. Add the server

Claude Code, one command:

claude mcp add gemini-relay -- npx -y gemini-relay

Claude Desktop, Cursor, Windsurf and a global npm install are in Installation.

Then run /mcp inside Claude Code. Eleven tools should be listed, all prefixed gemini-: gemini-ask, gemini-plan, gemini-image, gemini-models, gemini-doctor, gemini-brainstorm, gemini-fetch-chunk, gemini-conversations, gemini-cancel, gemini-ping, gemini-help.

3. Check it worked

Ask your agent:

"Run gemini-doctor"

It reports whether the CLI was found, whether you are signed in, and how much quota is left. It costs nothing to run. A healthy report ends with:

🎉 **System Ready!** The modern Antigravity CLI backend is active, signed in, and operational.

From a clone of this repository, npm run doctor runs the same check from the terminal.

If it says agy was not found even though your own shell finds it, the server did not inherit your PATH — see when the server cannot find agy. If the login check fails, run agy once interactively to sign in.

For AI agents — the exact checks and floors

Node floor	18.19.0, declared in package.json engines. CI runs 18.x, 20.x and 22.x.
Tool count	Nine. A tenth, gemini-timeout-test, is registered only when GEMINI_MCP_TEST_TOOLS is set — see environment variables.
gemini-doctor reports	The active backend and where that choice came from; the resolved agy and gemini executables and their versions; and — when agy is the active backend and was found — login and quota.
How it reads quota	By running agy -p "/usage" --output-format json. Agy's own command layer answers that without an agent turn, so it costs no tokens.
npm run doctor	The same diagnostic, run from a clone of the repository rather than through MCP.

Next

Run your first calls in First Steps.