How to clip videos from Claude, ChatGPT, Gemini CLI and Cursor with Katto

Katto turns a long video into short vertical clips: it finds the moments worth posting, cuts them to 9:16, reframes around whoever is on screen, burns in captions and gives every clip a score from 0 to 100 with the reason. You can do all of that from the Katto dashboard, but also from the AI assistant you already have open. This guide shows how to connect Katto to Claude, ChatGPT, Claude Code, Gemini CLI and Cursor, and what to ask once it's connected.

What you need

A Katto account: the free plan gives you 2 videos a month, up to 30 minutes each, no card needed. Katto's server is hosted at https://mcp.katto.tech/mcp, so there is nothing to install, and you sign in with your Katto account through OAuth: the assistant never sees your password or an API key.

Claude (web and desktop)

Open Settings → Connectors → Add custom connector, name it Katto, paste https://mcp.katto.tech/mcp, then sign in to Katto on the page that opens and allow access. Katto is also submitted to Anthropic's Connectors Directory; until the listing is approved, the custom connector gives you the same tools.

ChatGPT

Go to chatgpt.com/plugins, click +, choose Add custom MCP server, enter https://mcp.katto.tech/mcp with OAuth authentication, then sign in to Katto when asked. Katto is submitted to the ChatGPT plugin directory as well, and is in review there.

Claude Code

claude mcp add --transport http katto https://mcp.katto.tech/mcp

Then run /mcp in Claude Code and sign in when it asks.

Gemini CLI

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

Then /mcp auth katto to sign in. If you moved to Antigravity CLI, the Gemini setup page shows the equivalent configuration.

Cursor, VS Code and other MCP clients

Add the hosted server to your client's MCP configuration (for Cursor, ~/.cursor/mcp.json or .cursor/mcp.json in a project):

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

Prefer a local server? npx -y katto-mcp with an API key from your Katto dashboard does the same job.

What to ask

• "What Katto plan am I on, and how many videos do I have left this month?"
• "Turn this podcast into short clips and tell me which one to post first: [YouTube link]". Katto processes the video in its cloud, which takes a few minutes for a 20-minute source; the assistant follows the job and shows each clip with its title and score.
• "Re-render the second clip with the split-screen layout". Edits like a new layout, a trim or translated captions don't use any video quota.
• "Get the transcript of my latest video".
• "Write a TikTok caption for the best clip, then schedule it for tomorrow at 6 pm". Publishing is marked as an irreversible action, and Katto tells the assistant to get your explicit go-ahead before posting.

There are 29 tools in total, from creating a job or uploading a local file to retitling a clip, adding a hook, translating captions and publishing to 7 platforms. The full list says what each one does and what it costs in quota.

Things worth knowing

Quota works the same as on the site: creating a job uses one video of your monthly allowance, edits don't. Long videos take minutes, not seconds; you can close the conversation and find the clips in your dashboard, and you get an email when they're ready. Free clips are kept 3 days; on a paid plan they're never deleted.

One honest note: from September 30 to October 19, 2026, tool calls on the hosted connector failed with an "HTTP 404" error. A security change on our side blocked the connector's own requests while the tool list still loaded, so it looked connected. It's fixed. If you tried Katto from an assistant during those weeks and gave up, it's worth another go, and you don't need to reconnect.

Pricing

Katto is billed per source video, not per minute or credit. Free: 2 videos a month up to 30 minutes. Solo: 10 videos up to 60 minutes. Creator: 25 videos up to 90 minutes, with AI dubbing in 19 languages and scheduling. Assistant access is included on every plan. More on Katto's MCP server.