Clip from a prompt — everywhere you work
Give Katto a long-form video and a goal. It finds the strongest moments, writes captions, reframes to 9:16, and schedules — from the app, your terminal, or any AI assistant. A tool needs a person clicking in a dashboard; an agent takes a plain-language goal from a chat, a script, or an AI assistant. Same agent, same quota, everywhere.
“Clip yesterday’s webinar, add captions, and schedule 3 shorts this week at 9am.”
→ First clip in ~4–5 minutes. No manual selection.
One agent, three surfaces
Connect Claude, Cursor, ChatGPT, VS Code or Windsurf to the hosted MCP server. One-click OAuth 2.1 — no key on disk — or an sk_live_ key on the same URL.
{ "mcpServers": { "katto": { "url": "https://mcp.katto.tech" } } }MCP server →Run the whole pipeline from your shell or a cron. Submit, poll, fetch clips — with --json for scripts and an idempotency key on every submit.
npm i -g katto-cli katto clip https://youtu.be/… --waitRead the docs →
A versioned /v1 API with async enqueue, scoped keys, idempotency and HMAC-signed webhooks. Build Katto into any workflow.
POST /v1/jobs { "url": "https://youtu.be/…" }API reference →What the agent runs
Videos up to 3 hours · YouTube, Twitch, Vimeo, Rumble, Zoom, Dailymotion, or an upload.
Built for agents that ship
Point your agent at Katto
Free to start. The same account works in the app, CLI, MCP and API.