hwatu
Lets a coding agent actually prove a UI change worked — by opening a prewarmed browser window in ~13ms and scoring the pixel difference — instead of the agent just claiming the fix looks right based on the code it wrote.
🔗 Visit hwatu
Description
A coding agent that changes CSS or a component's layout has no real way to know whether the change looks right — it can read the code it wrote, but it can't see the rendered result the way a human reviewer would. hwatu closes that gap: it's a background WebKitGTK daemon that keeps a browser view prewarmed and ready (already loaded, GPU compositor hot), so opening a fresh window takes a median of 13ms instead of the usual multi-second browser startup.
Once open, it doesn't just take a screenshot — it produces pixel-difference scoring with heatmaps, and quantifies animations as concrete metrics (duration, easing, velocity) rather than a subjective "looks smooth" judgment. It connects to agents via an MCP server, ships with a built-in ad blocker using 119k EasyList rules, and — because it's genuinely a working browser, not just a screenshot tool — doubles as a fast daily browser for people using tiling window managers.
💬 Our review
The short version: if you're running coding agents that touch UI code and want a fast, deterministic way for them to verify visual changes instead of trusting their own claims, hwatu is a free, open-source tool built exactly for that — reportedly 9x faster than Playwright for this specific use case.
Compared to Playwright or Puppeteer — the standard tools for browser automation and visual regression testing — hwatu isn't trying to replace full end-to-end test suites; it's optimized narrowly for fast, repeated verification loops during active agent-driven development, where speed matters more than comprehensive test coverage. The tradeoff is scope and platform: it requires the WebKitGTK system library (a Linux-centric dependency), and as a young project it doesn't have the ecosystem, browser coverage, or track record of an established framework like Playwright.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Gratuit et open-source (AGPL-3.0)
Pros
Ouverture de fenêtre en ~13ms grâce au préchauffage
Scoring de différence pixel avec heatmaps, pas juste une capture
Intégration MCP directe pour les agents IA
9x plus rapide que Playwright pour la vérification visuelle rapide
Cons
Dépendance système WebKitGTK, plutôt Linux-centrique
Projet jeune, écosystème et couverture navigateurs limités vs Playwright
Pas un remplaçant de suite de tests E2E complète