Independent benchmark · v0.1
Which AI design add-ons actually make better UI?
We run Claude Code, Cursor and Codex skills, DESIGN.md files and design MCP servers against one standard brief and score the output on design-token use, accessibility, consistency and generic “AI slop”. Same brief, same tokens, published transcripts.
Leaderboard
Items we haven’t run yet are shown as Untested — we never estimate or borrow scores. Default order is GitHub stars (popularity, not quality); stars of a shared collection repo are shown but not ranked.
| Agents (per docs) | License | ||||
|---|---|---|---|---|---|
| UI UX Pro Maxnextlevelbuilder | Skill | Untested | Claude Code, Cursor, Codex, Gemini CLI, Windsurf, GitHub Copilot, OpenCode, Kiro, Antigravity | 142.0k | MIT |
| Awesome DESIGN.md (VoltAgent)VoltAgent | DESIGN.md | Untested | Any (generic), Google Stitch | 124.2k | MIT |
| Taste Skill (design-taste-frontend)Leonxlnx | Skill | Untested | Claude Code, Codex, Cursor | 104.3k | MIT |
| ImpeccablePaul Bakaus (pbakaus) | Skill | Untested | Claude Code, Cursor, Codex, Gemini CLI, GitHub Copilot, OpenCode, Kiro, Antigravity | 92.4k | Apache-2.0 |
| Chrome DevTools MCPChrome DevTools (Google) | MCP server | Untested | Claude Code, Cursor, Gemini CLI, GitHub Copilot, VS Code, Antigravity | 54.4k | Apache-2.0 |
| Emil Kowalski's Skills (emil-design-eng, animate…)Emil Kowalski | Skill | Untested | Any (generic) | 51.1k | MIT |
| Playwright MCPMicrosoft | MCP server | Untested | Claude Code, Cursor, Codex, VS Code, Windsurf, Gemini CLI, Any (generic) | 38.8k | Apache-2.0 |
| DESIGN.md format spec (Google Labs)Google Labs | DESIGN.md | Untested | Google Stitch, Any (generic) | 28.6k | Apache-2.0 |
| Framelink Figma MCP (Figma-Context-MCP)Graham Lipsman (GLips) | MCP server | Untested | Cursor, Any (generic) | 16.1k | MIT |
| UI SkillsJulien Thibeaut (ibelick) | Skill | Untested | Any (generic), MCP clients (registry) | 10.3k | MIT |
| Talk to Figma MCPGrab | MCP server | Untested | Cursor, Claude Code, Codex | 7.1k | MIT |
| Magic MCP (21st.dev)21st.dev | MCP server | Untested | Cursor, Claude Code, Windsurf, VS Code | 6.3k | ISC |
| Interface DesignDammyjay93 | Skill | Untested | Claude Code, Codex | 5.9k | MIT |
| Excalidraw Diagram SkillCole Medin (coleam00) | Skill | Untested | Claude Code, OpenCode | 5.1k | Not stated |
| Awesome Design Skills (TypeUI)Bergside | DESIGN.md | Untested | Claude Code, Cursor, Codex, Google Stitch | 3.4k | MIT |
| Designer Skills Pack (Owl-Listener)Owl-Listener | Skill | Untested | Claude Code, Gemini CLI | 3.1k | MIT |
| Figwrightawdr74100 | MCP server | Untested | Claude Code, Cursor, Codex, Any (generic) | 1.2k | MIT |
| Designer Skills (Julian Oczkowski)Julian Oczkowski | Skill | Untested | Claude Code, Cursor, GitHub Copilot, Codex | 594 | Apache-2.0 |
| brands-design-mdricocc | DESIGN.md | Untested | Any (generic) | 160 | Not stated |
| AccessLint a11y skillsAccessLint | Skill | Untested | Claude Code, Cursor, Codex, GitHub Copilot, Windsurf, Gemini CLI, Cline | 114 | MIT (per README) |
| Mobbin MCPMobbin | MCP server | Untested | Claude, ChatGPT, Codex, Cursor, Antigravity | 64 | MIT (config repo; service is paid) |
| design-md catalog (WebDesignHot)WebDesignHot | DESIGN.md | Untested | Any (generic), Claude, Cursor, Cline, GitHub Copilot | 3 | MIT |
| tokensmithjtmchorse | MCP server | Untested | Claude Code, Cursor, Claude Desktop | 1 | MIT |
| Gummble MCPGummble | MCP server | Untested | Claude Code, Cursor, Codex, Windsurf, Claude Desktop | 0 | MIT (config repo; service is paid) |
| @sorb/tapmetatoy | MCP server | Untested | Claude Code, Cursor, Any (generic) | 0 | Apache-2.0 |
| frontend-design (Anthropic)Anthropic | Skill | Untested | Claude Code | 184.5k stars: collection repo (anthropics/skills) | Apache-2.0 |
| theme-factory (Anthropic)Anthropic | Skill | Untested | Claude Code | 184.5k stars: collection repo (anthropics/skills) | Apache-2.0 |
| figma-implement-design (OpenAI × Figma)OpenAI / Figma | Skill | Untested | Codex | 28.4k stars: collection repo (openai/skills) | Figma Developer Terms (beta) |
| figma-create-design-system-rules (OpenAI × Figma)OpenAI / Figma | Skill | Untested | Codex | 28.4k stars: collection repo (openai/skills) | Not stated in skill folder |
| web-design-guidelines (Vercel)Vercel Labs | Skill | Untested | Any (generic) | 33.2k stars: collection repo (vercel-labs/agent-skills) | MIT |
| Figma MCP server (official)Figma | MCP server | Untested | Claude Code, Cursor, Codex, VS Code, GitHub Copilot, Gemini CLI, Kiro, Warp, Xcode | 2.1k stars: guide repo (figma/mcp-server-guide) | Proprietary service (guide repo: no license file) |
| shadcn MCPshadcn | MCP server | Untested | Claude Code, Cursor, VS Code, Codex, OpenCode | 126.8k stars: host repo (shadcn-ui/ui) | MIT |
| Google Stitch MCPGoogle | MCP server | Untested | Any (generic) | — | Proprietary service |
| Storybook MCP (@storybook/addon-mcp)Storybook | MCP server | Untested | Claude Code, Codex, VS Code | 271 stars: repo (archived — moved into storybookjs/storybook) | MIT |
| Penpot MCP server (official)Penpot | MCP server | Untested | Claude Code, Claude Desktop, Any (generic) | — | MPL-2.0 |
| Supernova Consumer MCPSupernova | MCP server | Untested | Any (generic) | — | Proprietary service |