← All recipes Every week
AI QA tester
Every week, a brand-new customer tries your app and reports every bug they trip over.
An agent signs up, does the core job end to end in a real browser, then goes looking for trouble, and files what it finds with screenshots.
How it runs
-
It starts
Once a week, a task starts on its own.
-
The agent works
Starts your app with realistic data and uses it in headless Chrome like a first-day customer: empty states, odd inputs, the back button, a phone screen.
-
You approve
Read the illustrated report, then start the bug tasks worth fixing.
The prompt
Paste it into the automation’s instructions. Each run starts a task from it.
Start the app with realistic data and use it in headless Chrome the way a new customer would on their first day. Get set up and do the core job end to end, then wander: empty states, odd inputs, the back button, a phone-sized screen.
For every real bug you hit — something broken, wrong or confusing, not a matter of taste — capture a screenshot, the exact steps to reproduce, and what you expected instead. Check the backlog for duplicates, then file one backlog task per bug (don't start them). File at most five, worst first.
Attach qa-report.html: your route through the app with a screenshot at each step, and the bugs you filed marked where you hit them. Don't change any code.Set it up
In your project, open Settings → Automations → New automation.
- Trigger
- Schedule: weekly
- Action
- Create task
- Start each task
- One-shot implementation