Back to Kontent.ai Projects
🔄
🚀 Hackathons

Design System Comparison AI Tool (Figma Plugin Implementation)

Kontent.ai

AI-powered Figma plugin that automatically compares design system components with documentation – reducing manual comparison time from hours to minutes.

Design System Comparison AI Tool (Figma Plugin Implementation)

🎯 Problem

Manual comparison challenges with design system discrepancies.

  • Design systems have discrepancies between documentation (Kontent.ai) and implementation (Figma)
  • Manual comparison takes 2-3 hours per component
  • Missing features go unnoticed until production
  • New team members struggle to understand what's actually available
Design System Comparison Tool
Plugin interface showing component comparison
Comparison Results
Detailed comparison results
Analysis Progress
Visual progress feedback during analysis
Generated Report
Automated Confluence report generation

🔄 Development Journey

Three-phase evolution from exploration to production-ready plugin.

  • Phase 1: MCP Servers (Exploration) – Used Model Context Protocol servers for rapid API prototyping with Kontent.ai, Atlassian, and Figma. Validated that AI can effectively compare design systems semantically.
  • Phase 2: CLI Script (Automation) – Node.js command-line tool with direct API calls. Successfully automated the comparison but required terminal knowledge – limiting adoption.
  • Phase 3: Figma Plugin (Final Solution) – Native plugin working directly in Figma's interface with visual UI, making the tool accessible to non-technical designers.

🏗 Technical Architecture

Technology stack and integrations.

  • Frontend: TypeScript Figma Plugin with custom HTML/CSS UI
  • Backend: Vercel serverless functions (proxy for CORS)
  • AI: Anthropic Claude API for semantic comparison
  • APIs: Kontent.ai Management API, Atlassian Confluence API

Key Features

Advanced capabilities for design system validation.

  • Intelligent Comparison – Semantic matching recognizes that "showIcon" property = "Icon in headline" feature (not just exact string matching)
  • Category Weighting – Design tokens: 30%, Structure: 20%, Optional elements: 40%, Behavioral properties: 10%
  • Automated Reporting – Generates Confluence pages with coverage scores, gaps, and recommendations
  • Visual Progress – Step-by-step feedback during analysis so users understand what's happening

🔧 Technical Challenges Solved

Key technical problems overcome during development.

  • CORS restrictions → Vercel backend proxy
  • Browser API limitations → Custom base64 encoding for Figma environment
  • State management between plugin UI and code
  • Secure API key storage via clientStorage

🌊 Impact

Time Savings: 2-3 hours → ~1 minute per component. AI catches semantic matches humans might miss. Automated reports for stakeholders. Run on every design update to catch drift early.

Vlado Krejci | Senior Product Designer