Figma Design Reviewer
AI-Powered Design QA
A web app where three AI agents review Figma designs in parallel — checking UX heuristics, visual polish, and accessibility — then let you post approved feedback back as Figma comments.
- Role
- Designer & Developer
- Team
- Solo project with AI collaboration
- Timeframe
- 2026
Overview
Figma Design Reviewer is a multi-agent workflow that turns a selected Figma frame into structured design feedback. I built it to test a hunch: that AI critique gets more useful when it’s grounded in the actual artifact and connected to the tool where design teams already work.
How It Works
The prototype connects to Figma via OAuth, analyzes a selected frame, and starts with a context-analysis step that infers likely product type, user intent, and review priorities. Before review begins, you can confirm, edit, or skip those assumptions.
From there, specialized AI reviewers run in parallel — UX heuristics, visual polish, accessibility and edge cases. Their outputs get normalized into structured issues, mapped back to relevant nodes, and surfaced in the app for filtering, review, and selective posting back into Figma as comments.
Most AI feedback tools stop at broad commentary. I wanted something more operational — critique that interprets the artifact, separates feedback into clear lenses, keeps you in the loop, and returns output you can actually act on.
On the design of the review workflow
Why It Matters
What interested me most wasn’t the model output — it was the workflow design. The project shows how agentic systems can be orchestrated around real tools and real decisions, not just prompts.
