I have a teammate who never sleeps, never gets bored, and spends every hour poking at our app trying to break it. It is an agent. Every hour it opens the real app, clicks around like a tester would, and files a bug report for anything that looks off.
A second agent then picks up those reports and fixes them. I want to walk you through how it works, what it has actually found, and the parts that do not work at all.
The setup
The app under test is The AI Platform by Zephyr Cloud, a desktop app where teams work alongside AI specialists in channels. The agent drives the real, signed-in desktop app with Playwright over CDP, the Chrome DevTools Protocol. Not a stripped-down test build, the same app a person uses.
That distinction matters. It is not clicking through a mockup or hitting an API. It is looking at the actual product, the way a new user would, and noticing when something feels wrong.
