58 automated rules cover contrast, layout, typography, heuristics, and color — each scored, graded A–F, and fixable in one click. Get a second opinion from 15 AI models.
Top findings
Features
Every check is documented. Contrast ratios computed with alpha compositing. Spacing verified against 8pt grids. Each rule references a specific WCAG criterion or Nielsen principle — no guesswork.
Applied directly on canvas. Verified before the issue closes. Undo any change instantly.
Scored 0–100, weighted by severity. Errors count 3× more than warnings.
Bring your own API key. Claude, GPT-4o, Gemini, and 12 more run in parallel — results attributed per provider.
Register your component library. Detached instances, drift from canonical size, unregistered components.
Full audit as structured JSON or formatted PDF. Built for handoff, post-mortems, and design reviews.
How it works
Choose any frame or your entire page. Design Buddy works at any scope — a single component or a full-page mockup.
58 rules run in parallel. Contrast computed with alpha compositing. Spacing checked against 8pt grids. Results scored by severity and graded instantly.
One click applies and verifies each fix directly on canvas. The score refreshes live. Export as JSON or PDF for handoff.
Audit categories
Every audit runs all six evaluators in parallel, then combines them into a single weighted grade.
Text legibility against WCAG AA/AAA — including opacity, gradients, and colour-blind safety.
Spacing grids, alignment, padding symmetry, off-canvas elements, and minimum touch-target sizes.
Font sizes, line height, hierarchy, weight contrast, and typeface variety across the design.
Nielsen's 10 usability principles checked algorithmically — feedback, control, consistency, error prevention.
Palette size, near-duplicate colours, saturation balance, and colour-only information encoding.
Component coverage, unregistered instances, detached components, and size drift from registered specs.
AI analysis
All analysis uses your API key, runs inside Figma, and never touches a third-party server.
Ollama runs entirely offline — no API key required.
Scoring system
Scores use a damped sigmoid formula so a design with a few warnings lands around 70–80, not near zero.
The overall score is a weighted average across all 6 categories: contrast at 30%, layout/typography/heuristics at 20% each, color at 10%.
Pricing
$0/month
Start auditing immediately — no credit card, no account.
—
Unlimited audits across all your devices.
Billed monthly via Lemon Squeezy · Cancel anytime
FAQ
Never. All evaluation logic runs inside Figma's UI iframe — your design data never leaves Figma. When you enable AI analysis, only a structured text summary (not raw node data) is sent to your chosen AI provider, using your own API key.
You get 10 complete audits every 30 days, per device. No signup, no account — just install and run. Cancelled or errored audits don't count against your quota. The counter resets on a rolling 30-day window.
15 providers: Anthropic Claude, OpenAI, Google Gemini, Groq, OpenRouter, Azure OpenAI, GitHub Models, Mistral, Together AI, DeepSeek, xAI/Grok, Cerebras, Fireworks AI, Perplexity, and Ollama. Ollama runs fully offline — no API key required.
Yes — Pro licenses support multiple device activations. The exact limit is shown at checkout. Licenses also include a 7-day offline grace period, so your audit access stays available if you temporarily lose connectivity.
Yes. Design Buddy walks the full node tree, resolving INSTANCE nodes to their main components. Library compliance checks specifically target instances and detect detached components by name. Up to 10,000 nodes per audit.
Each category is scored 0–100 using a damped sigmoid: errors count 12 points each, warnings 4, info 1. The overall score is a weighted average — contrast at 30%, layout/typography/heuristics at 20% each, color at 10%. Design Library compliance optionally takes a 10% slice when enabled.