Use Katto with Gemini

Katto’s official MCP server turns long videos into scored, captioned 9:16 clips from Gemini CLI or its successor, Antigravity CLI. It’s hosted: you add one URL (or one extension), sign in to Katto in your browser, and nothing is stored on disk.

Gemini CLI

Install the Katto extension from its public repository (needs git):

gemini extensions install https://github.com/miracleweasel/katto-mcp

Then run /mcp auth katto inside Gemini CLI to sign in. The extension only declares the hosted server and a short context file telling Gemini how to use the tools and the quota; it contains no key.

Or by hand, in ~/.gemini/settings.json:

{
  "mcpServers": {
    "katto": { "httpUrl": "https://mcp.katto.tech/mcp" }
  }
}

httpUrl selects the Streamable HTTP transport. Gemini CLI discovers the OAuth configuration from the server on its own.

Antigravity CLI

Google is moving Gemini CLI users to Antigravity CLI (agy). Add the hosted server to ~/.gemini/config/mcp_config.json (or .agents/mcp_config.json for one workspace). Antigravity uses serverUrl where Gemini CLI used httpUrl:

{
  "mcpServers": {
    "katto": { "serverUrl": "https://mcp.katto.tech/mcp" }
  }
}

Already installed the Gemini CLI extension? agy plugin import gemini converts your existing extensions and MCP servers.

Gemini app

Katto isn’t in the Gemini app’s built-in connector list: Google picks those partners itself. Some accounts can add a custom app by URL at the bottom of gemini.google.com/apps, starting with Gemini Spark. If your account has that option, paste https://mcp.katto.tech/mcp. We haven’t been able to test this path on every account type, so the CLI setups above are the ones we support.

Then just ask

Clip the 3 best moments from
https://www.youtube.com/watch?v=… into 9:16 with
word-timed captions, then give me the download links.

Gemini calls katto_create_clip_job, polls until it’s done, and returns the MP4 + caption URLs. Every tool is documented on the tools reference. One clip job uses 1 video from your monthly quota (25 on Creator, 2 on Free); re-renders are free, dubbing is a Creator feature.

Troubleshooting