Frontend QA that shows its evidence
Give us a URL.
We hand back broken journeys — with proof.
buildx opens your deployed site like a real visitor: clicks the buttons, fills the forms, watches the cart. When a user-visible interaction contradicts what the page promises, it records the steps, the screenshots and the exact moment of failure.
shop.example/cart
Expectedenabled “Remove” button removes the line item and updates the total
Observedclick accepted; cart still shows 2 items, total unchanged after 3s
- 1. navigate to /cart
- 2. observe → “Remove” (button, enabled)
- 3. click [e7]
- 4. wait 3s; assert_text “1 item” → not found
How a run works
Step 1
Describe the target
A URL is enough. Optionally add one workflow to focus on, the outcome you expect, and disposable test credentials. You watch every step and can stop the run at any time.
Step 2
The agent investigates
A remote model drives a real browser through native clicks and typed assertions — no DOM shortcuts that would pass a test a human could not complete.
Step 3
Read the report
Per-scenario verdicts, evidence and reproduction steps, plus what was blocked or untested. A bounded run never claims “the whole site passed”.