Hi HN, I built Behavr - https://behavr.ai/ - it runs AI agents through your Figma prototypes to find UX issues in minutes, providing quantitative and qualitative UX insights.
The Problem:
Most product teams either skip UX research entirely or spend weeks (even months) testing. This is expensive and time consuming. Behavr offers UX insights in minutes, allowing you to iterate and test as fast as you design.
How it works:
Paste your Figma URL, Behavr spawns AI users with diverse personas, they navigate your prototype like real users, you get UX insights in under 3 minutes.
What makes the AI users realistic:
Each agent gets a profile, motivation (focused completer, casual browser, skeptical evaluator), tech literacy, age, and gender — these all affect patience and behaviour
Effort budgets calibrated to task complexity. A casual browser bails on a 7-screen signup flow, a focused completer won't
Grounded in UX research: F-pattern scanning, choice overload, Fitts's Law, decision fatigue. Behavioural priors drawn from published studies by Nielsen Norman Group, Baymard Institute, and Google UX Research
Issues mapped to Nielsen's 10 usability heuristics
Under the hood:
Figma API extracts screens and interactive element maps
Claude Sonnet handles vision analysis and agent simulation
Parallel execution via Celery
Outlier detection and hallucination checking before synthesis
Hi HN, I built Behavr - https://behavr.ai/ - it runs AI agents through your Figma prototypes to find UX issues in minutes, providing quantitative and qualitative UX insights.
The Problem: Most product teams either skip UX research entirely or spend weeks (even months) testing. This is expensive and time consuming. Behavr offers UX insights in minutes, allowing you to iterate and test as fast as you design.
How it works: Paste your Figma URL, Behavr spawns AI users with diverse personas, they navigate your prototype like real users, you get UX insights in under 3 minutes.
What makes the AI users realistic: Each agent gets a profile, motivation (focused completer, casual browser, skeptical evaluator), tech literacy, age, and gender — these all affect patience and behaviour Effort budgets calibrated to task complexity. A casual browser bails on a 7-screen signup flow, a focused completer won't Grounded in UX research: F-pattern scanning, choice overload, Fitts's Law, decision fatigue. Behavioural priors drawn from published studies by Nielsen Norman Group, Baymard Institute, and Google UX Research Issues mapped to Nielsen's 10 usability heuristics
Under the hood: Figma API extracts screens and interactive element maps Claude Sonnet handles vision analysis and agent simulation Parallel execution via Celery Outlier detection and hallucination checking before synthesis
Stack: Next.js, FastAPI, Celery + Redis, PostgreSQL, Claude Sonnet, Vercel + Railway.
Free beta open now. 3 tests, no credit card.
Would love feedback from anyone who designs or builds products. Thanks!