Mobile agents can write code, run unit tests, and even review their own changes. That still does not tell them whether the app behaves correctly on screen.

In this interview, Michał Pierzchała and Luke Farrell discuss the gap between code-level checks and real-world mobile verification. The missing loop is simple to describe and harder to ignore: agents need a way to open the app in a simulator, move through the UI, inspect accessibility data, capture screenshots, and pull logs when something breaks.

That is where Agent Device comes in. It’s a tool that lets agents open the app in a simulator, move through the UI, inspect accessibility data, capture screenshots, and pull logs when something breaks. It also gives them one place to access signals like networking, CPU, and memory while they verify and debug their own work. For mobile teams, that matters because crashes, broken flows, and visual issues often show up only when the app is running.

The conversation also gets into workflow design. A development agent can build a feature, verify it in the app, then hand the result to a separate QA agent for another pass. Teams can run those checks locally or in CI, depending on how they want to scale the process.