design-token-bridge-mcp

Verified on AgentGraph — trust infrastructure for AI agents. View profile

An MCP server that translates design tokens between platforms. Extract tokens from Tailwind, CSS, Figma, or W3C DTCG format — then generate native themes for Material 3 (Kotlin), SwiftUI (with Liquid Glass), Tailwind, and CSS Variables.

Built for the v0 → Figma → Claude Code design pipeline.

┌──────────────┐     ┌──────────────┐     ┌────────────────────────┐
│  Tailwind     │     │              │     │  Material 3 (Kotlin)   │
│  CSS Vars     │────▶│  Universal   │────▶│  SwiftUI (Swift)       │
│  Figma Vars   │     │  Token       │     │  Tailwind Config       │
│  W3C DTCG     │     │  Schema      │     │  CSS Variables         │
└──────────────┘     └──────────────┘     └────────────────────────┘
    Extractors           Bridge              Generators

Install

From npm

npm install -g design-token-bridge-mcp

From source

git clone https://github.com/kenneives/design-token-bridge-mcp.git
cd design-token-bridge-mcp
npm install && npm run build

Configure with Claude Code

Add to your Claude Code MCP settings (~/.claude/settings.json or project .mcp.json):

{
  "mcpServers": {
    "design-token-bridge": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "design-token-bridge-mcp"]
    }
  }
}

Or if installed from source:

{
  "mcpServers": {
    "design-token-bridge": {
      "type": "stdio",
      "command": "node",
      "args": ["/path/to/design-token-bridge-mcp/build/index.js"]
    }
  }
}

Tools (9 total)

Extractors — Input → Universal Tokens

Generators — Universal Tokens → Output

Validation

Universal Token Schema

All tools speak this common format:

{
  "colors": {
    "primary": { "value": "#6750A4", "description": "Brand primary", "category": "primary" }
  },
  "typography": {
    "display-large": { "fontSize": 57, "lineHeight": 64, "fontWeight": 400, "fontFamily": "Inter" }
  },
  "spacing": { "xs": 4, "sm": 8, "md": 16 },
  "radii": { "sm": 8, "md": 12, "lg": 16 },
  "elevation": {
    "low": {
      "shadowColor": "#000000",
      "shadowOffset": { "x": 0, "y": 2 },
      "shadowRadius": 4,
      "shadowOpacity": 0.1
    }
  },
  "motion": {
    "fast": { "duration": 150, "easing": "ease-out" }
  }
}

Example: Full Pipeline

1. Extract tokens from a Tailwind config

Use extract_tokens_from_tailwind with the contents of my tailwind.config.js

2. Generate native themes from the extracted tokens

Take those tokens and run them through:
- generate_tailwind_config (for the web app)
- generate_material3_theme (for Android)
- generate_swiftui_theme with liquidGlass=true (for iOS)
- generate_css_variables (for a vanilla CSS fallback)

3. Validate accessibility

Run validate_contrast on those tokens at AAA level

Example Output

See the examples/qt-games/ directory for a complete responsive landing page built entirely from MCP-generated tokens, including:

• qt-games-tokens.json — extracted universal tokens
• tailwind.config.js — generated Tailwind config
• variables.css — generated CSS custom properties
• contrast-report.json — WCAG validation (AAA pass)
• index.html + styles.css — responsive landing page using the generated tokens

v0 + Figma Free Tier Setup

This MCP works with free tiers of both v0 and Figma. See the setup guides:

• v0 Setup Guide — free tier signup, web UI workflow
• Figma Setup Guide — free tier MCP config (6 calls/month)
• Claude Code Setup — full pipeline configuration
• Token Extraction Guide — manual extraction when APIs aren't available

Tech Stack

• TypeScript + Node.js
• @modelcontextprotocol/sdk v1.x (stdio transport)
• Zod for schema validation
• Zero heavyweight dependencies — no Babel, no PostCSS, no Style Dictionary

Tests

# 91 unit tests
npm test

# 31 Playwright visual/responsive tests
npm run test:e2e

Contributing

See CONTRIBUTING.md for development setup, project structure, and PR guidelines.

License

MIT — see LICENSE.