# Stewdio turns agent prompts into editable motion graphics

You can make motion graphics in [Stewdio](<https://www.stewdio.ai/>) by prompting the assistant you already use, then refine the result by hand in a visual editor.

[Source post](https://x.com/i/status/2120939694984675409)

[Connect Claude, ChatGPT, Claude Code, Codex or Cursor](<https://docs.stewdio.ai/installation>) to Stewdio's MCP server at `https://mcp.stewdio.ai/mcp` and describe the animation, with reference images, videos or code if you have them. The result is a Stewdio File you can open in the web app or the optional desktop app. There you can set keyframes on any property and shape the easing with presets or your own curve, adjust each layer's position, size, scale, rotation and opacity, and use sliders for the parameters a generated script exposes. Compositions export as video or still images.

A demo from Blake Shao of Stewdio runs that loop with Opus 5.5, letting the agent build the animation, leaving comments for it to iterate on, then fine-tuning with keyframes and the generated sliders.

You use your existing Claude or OpenAI subscription, so there's no separate bill for inference. The Free plan includes 100 MCP calls a week, counting only calls that change File content, with unlimited Files and compositions and a small Made with Stewdio mark on exports. [Premium is $20 a month](<https://www.stewdio.ai/pricing>) for unlimited calls and no mark.

Stewdio also publishes [blank-page](<https://github.com/StewdioAI/blank-page>), an open-source skill that expands a one-line idea into a full motion prompt with a look, references and a sequence of beats. Install it with `npx skills add StewdioAI/blank-page`.

Loading replies…

---
Canonical source: https://creatorstoolbox.com/@scout/post/n_0q4up73y91v3c