Hi Product Hunt, I'm the maker of Mlola UI. My coding agent could build a good screen from day one. What I couldn't get was the hundredth screen looking like the first: padding, colors, radius and gaps drifted a little every time, however carefully I prompted. So I built the system I wanted the agent to work inside. Spacing, type, radius and control heights come from scales, a gate fails any number typed into a component, and the behavior (layers, Escape, focus, the phone keyboard) ships with the components instead of being rebuilt per screen. For agents, `npx mlola-ui init` writes AGENTS.md, a design guide from your stylesheet and an MCP config. The agent looks components up, and check_markup flags a color or spacing it typed by hand. It works without an agent too. What's in it today: - 108 components: 54 free under MIT, 54 Pro (AI chat, agents, voice, code, canvas, editor, workflow, charts) - 25 blocks, 10 pages and 14 templates in Pro - Five themes from one spec, with WCAG 2.2 AA contrast solved by the theme engine - 1,314 browser tests in Chromium, Firefox and WebKit Pro is $99 once for one developer. The free core is enough to try the whole idea. I'd love to hear where it breaks for you, especially on a phone.

Mlola UI

The UI system your team and your agents build with

The UI system your team and your agents build with