Motion components your AI agent can use correctly.
Twenty-four motion, canvas and WebGL components for React, measured in real browsers, with a skill file that tells coding agents how to use them.
Born from friction.
Twenty-four parts. One kit.
Backgrounds, text reveals, image effects and scroll motion. They share one frame loop and one WebGL budget, so they work together instead of fighting.
Technique: Stone Age cave painting. Each person is painted in two poses, and the torch flicker swaps between them, the way the first animations worked.
Start with one command.
Add a component the way you add any shadcn component. It lands in your project as source code you can read and change.
Technique: red-ochre rock art seen by torchlight. Your cursor is a torch. Light a fire and your hand stencil is sprayed onto the rock.
Your agent picks the right part.
A skill file and llms.txt tell coding agents which component fits, which props exist, and what each one costs to run.
Technique: leather shadow puppets on lamp-lit cloth, in the spirit of Karnataka's Togalu Gombeyaata. The lamp moves behind each screen.
Measured, not promised.
Every component is tested in real browsers for frame time, layout shift and memory. Results on real phones are still pending, and the site says so.
Technique: early animation. Hand-inked drawings that boil at 12 frames a second, a hand-cranked projector, a film burn, then amber hand-tinting.
Small, and quick to start.
The runtime core is about 4 KB. WebGL loads only when a component is on screen and the device can handle it. MIT licensed.
Technique: the travelling ember burns the drawing into wood, like pyrography. The line cools from white-hot to a charred groove.