VoltAgent just released a library of 62 design system files formatted for AI coding agents. The project, called DESIGN.md and hosted on GitHub as "awesome-design-md," has 59.4k stars. Each file captures the visual language of a popular website. Stripe's gradients. Notion's minimalism. You drop one into your project, and coding agents like Cursor or Claude can build UIs that actually look like the brand you're referencing instead of the same generic output.

AI-generated interfaces tend to look identical regardless of what you're building. The design tokens, color palettes, typography specs, and spacing guidelines in these markdown files give agents enough context to produce something that feels intentional. The library covers everything from developer tools like Vercel and Supabase to consumer brands like Nike and Spotify. VoltAgent is clearly targeting the explosion of AI coding platforms, listing Lovable, Bolt, v0, and major LLMs as compatible tools.

But the project's commercial model has ruffled feathers.

CONTRIBUTING.md says pull requests for new designs can't be accepted "to maintain quality," pointing users instead to a paid request form. Community members have responded by opening GitHub issues asking for new designs, some publicly sharing their email addresses in confusion. A paid tier delivers light and dark theme files, HTML previews, and Discord access.

Selling brand-specific design files raises IP questions that haven't been answered. Raw design tokens like hex codes probably aren't copyrightable as functional facts, but the full "look and feel" of a brand's UI can constitute protectable trade dress under U.S. law. Repackaging closed-ecosystem designs like Apple's, which aren't publicly licensed, carries more risk than open-source systems like Vercel's. Using registered trademarks like "Spotify" or "Airbnb" to sell these files adds another legal layer. The project sits squarely in the messy territory courts are still working through, similar to the Getty Images versus Stability AI litigation. Claude Design suggests that as LLMs evolve, design is returning to code as the true source of design truth rather than proprietary systems like Figma.