Skip to content
Mixture Docs

What you can do with agents

Mixture is designed and built to work visually, with a powerful web-based canvas, and programmatically, via AI and agents. These pages are the programmatic side, written for you — a tour of what a connected agent can actually do, area by area, and what to ask for.

(Your agent doesn’t read these pages — it fetches its own, much drier documentation over its MCP connection the moment it connects. You never need to teach it anything.)

Two ways in:

  • Examples — walkthroughs of things people actually ask for, from putting text behind a person to starting a whole video from a song. Treat them as inspiration, not a menu: they compose (“the walkthrough, but localized, cut to this track”), and if you can describe an outcome, you can ask for it.
  • Capabilities (this section) — what’s possible in each area: motion, captions, data, video editing, formats and versions, design, effects, 3D, and how agents edit your project safely.

Everything an agent builds arrives as ordinary scenes, layers, and assets — whatever it makes, you can inspect and edit by hand in the same panels the canvas guides describe.