CHOOSEAgents & automation
Jev Automated Testing Agent
Jev Watch@jevwatch
Jev is an open-source agent that automates UI testing by selecting the next action based on the app's accessibility tree, offering a more efficient alternative to rigid scripts or full model inference per action.
TakeHow it works: agent-device reads the app’s accessibility tree (labels, buttons, fields—not screenshots), turns what’s on screen into allowed actions, and Jev chooses one.
Most automated tests either follow a rigid script or burn a full chat model on every tap. Callstack wired something in between: Jev picks the next UI action, and their open-source agent-device presses the button.
How it works: agent-device reads the app’s accessibility tree (labels, buttons, fields—not screenshots), turns what’s on screen into allowed actions, and Jev chooses one. Their demo run finished in 14 seconds for about $0.0023 in model inference.
Package: @typesafe-ai/sdk with model jev-latest. Fill actions already include the text to type—Jev selects the action, it doesn’t invent the string. Credit: Callstack (Mike Grabowski).