Cleya
Cleya is a creative-generation and video toolkit built for AI agents. It runs as a CLI, an MCP server, and a TypeScript library — the same 66 tools on all three surfaces.
It exists for one loop:
- Author an animation as declarative data — a scene spec, not hand-written HTML and timing code.
- Render it deterministically. Every frame is a pure function of time, so a preview frame is exactly the pixels the master will hold.
- Measure the rendered output with perception instruments — WCAG contrast per text box, optical alignment, dominant palette, brand conformance by ΔE.
- Correct on the numbers, deterministically, and promote what worked into a reusable starter.
Plenty of tools do step 1 and 2. Step 3 is the unusual one: Cleya reads its own output back as pixels and reports measurable facts about it, which is what lets an agent finish a job without a human squinting at every frame.
cleya draft --starter promo-hero --brand brand.json \
--copy '{"headline":"Ship it {accent:Friday}","tagline":"in one command"}'
That single call picks a guaranteed-valid skeleton, applies your brand tokens and copy, renders, then returns contrast and brand-conformance readings beside the file.
Where to go next
- Requirements — what must be on the machine
- Installation — clone, build, link, and wire up MCP
- Your first render — the loop end to end in five commands
- The scene DSL — the authoring grammar
Project status
Cleya is at v0.6.0 and actively maintained. 834 tests across 75 files, a documented scene grammar with a JSON Schema, and a contract-tested starter library — a change that breaks a starter breaks the build.
Licensed Apache-2.0. Issues and pull requests are welcome on GitHub.