iOS QA agent
A developer built a 'QA Engineer' agent for iOS using Maestro that analyzes code after tasks, auto‑generates visual tests to click and navigate the simulator, and self‑corrects via screenshot + coordinate estimation loops while keeping persistent script memory. (x.com). The post — showing the agent's workflow and persistence model — logged engagement (4 likes, ~2,369 views) as a practical take on automating UI QA. (x.com)
Everton Carneiro, who posts as @everton_dev and maintains iOS-focused repos and a YouTube channel, is the author behind the shared Maestro-based QA agent demo. (github.com/DevEverton) The agent implementation captures a screenshot plus the UI hierarchy each step, appends those observations to a running message history, and sends them to an LLM for next-action decisions — the Whitesmith writeup includes the exact loop and model calls used in their example. (whitesmith.co) Whitesmith’s example shows the agent parsing model responses into "tool_calls" and assistant text, then executing Maestro-style UI commands and checking for TASK_COMPLETED or TASK_FAILED signals to decide when to stop. (whitesmith.co) Maestro’s open-source repository, which has active commits and recent iOS-driver updates, provides the underlying commands and drivers the agent uses to run flows on simulators. (github.com/mobile-dev-inc/Maestro) Because Maestro’s official iOS support is primarily simulator-focused, community projects and guides have emerged to run Maestro flows on real iPhones; one community guide and repo detail a maestro-ios-device approach for physical-device execution. (dev.to) Commercial device clouds already integrate Maestro for broader coverage — BrowserStack documents running Maestro tests across more than 3,500 real iOS devices to supplement simulator runs. (browserstack.com)
