Audit what your AI actually built.
Drop a screenshot from Lovable, v0, Cursor or Bolt. Get a fidelity score, every off-brand colour, font and radius ranked by damage, and a repair prompt to paste straight back.
Drop a screenshot of the generated screen. You get ranked mismatches against this direction and a repair brief to paste straight back into your builder.
No evaluation yet. Findings appear here, worst first.
Questions
What is AI design drift?
AI builders like Lovable, v0, Cursor and Bolt gradually introduce off-brand colours, fonts, radii and spacing. Drift is the gap between your design system and what actually ships.
How does the audit work?
Upload a screenshot of what your AI built. A vision model compares it to your design tokens and anti-rules, scores fidelity 0–100, ranks every mismatch and writes a repair prompt.
Which builders does it support?
Any. The audit reads pixels, so it works on output from Lovable, v0, Cursor, Bolt, Claude, Replit, Windsurf or hand-written code.
Is it free?
You can define a design system and run audits with a free account. Pro ($15/mo) unlocks unlimited audits, full exports and MCP/API access for agents.