How we built a real iPhone device farm for iOS mobile testing

QA Wolf AI starts investigating failures while the suite still runs: it groups them by cause, repairs tests only with a passing run, and asks you when product intent is unclear.

QA Wolf, Claude Code, and Codex on the same 1,567 evals across web, iOS, and Android apps. QA Wolf passed 93%, Claude Code 78%, and Codex 74%.

Functional testing checks that software does what users expect. See the types, an example test, functional vs. non-functional, and what to automate first.

How QA Wolf runs Windows desktop app tests in parallel: single-use VMs from warm pools on Google Compute Engine, driven from Linux over CDP and Appium.

Building for MCP pushed QA Wolf to replace app navigation with focused, task-oriented views. See how designing for coding agents simplified the whole product.

We followed the same customers for 21 months and 50,000 AI-generated PRs: they deployed about 30% more often, reported bugs fell 20%, and bugs per deploy dropped by more than a third.

Claude Code and Codex can write a Playwright test. At hundreds of PRs a week, QA also needs coverage decisions, a test fleet, and agents that run unattended.

Agent sessions don't behave like cloud jobs. How we rebuilt our infrastructure so every QA Wolf AI agent gets its own machine with a real filesystem, shell and browser.