AnxForever/stylekit

What it solves

AI-generated websites often look generic because "make it look better" is too vague for a model to follow. StyleKit solves this by providing a structured library of visual styles—including design tokens, component recipes, and hard constraints—that AI agents can follow to produce consistent, high-quality visual designs without "style drift."

How it works

StyleKit transforms abstract visual directions into machine-readable specifications. It offers 148 curated styles across ten categories (such as Cyberpunk, Brutalism, and brand-inspired styles like Stripe or Apple). These styles are delivered to AI tools through three primary methods:

1. shadcn Registry: Styles can be added to existing projects via a CLI command to update CSS variables.
2. Agent Skills: A dedicated skill allows coding agents (like Cursor, Claude Code, or Windsurf) to apply specific styles and rules directly to the code.
3. MCP (Model Context Protocol): An MCP server provides style retrieval, design tokens, and a linting tool to verify if generated UI code adheres to the chosen style's rules.

Who it’s for

Developers and designers using AI coding agents to build front-end interfaces who want to move beyond generic AI aesthetics and implement specific, professional visual identities.

Highlights

• Extensive Library: 148 curated styles, 60 animations, and 36 page templates.
• Agent-Native: Deep integration with AI agents via MCP and Agent Skills for automated styling.
• Multi-Export: Supports Tailwind presets, shadcn themes, CSS variables, and Figma tokens.
• Machine-Readable: Includes /llms.txt and a documented HTTP API for easier AI consumption.
• Validation: Includes a stylekit_lint_code tool to ensure AI-generated code follows the style constraints.