FinalRun, a new AI‑driven mobile‑app testing tool, lets developers write tests in plain English and run them with vision‑based selectors across Android and iOS. The open‑source CLI installs with a single script, pulls Node.js and platform tools, and stores test artifacts in ~/.finalrun. By keeping tests in the repo, the framework avoids the drift that plagues external flow files.

Test definitions live in .finalrun/tests as YAML files, where each step is a natural‑language command. Secrets are resolved from workspace‑root .env files, keeping API keys out of version control. The CLI can execute a single spec, a suite, or launch a local report server, giving developers instant visual feedback without brittle selectors.

FinalRun’s vision‑based engine bypasses brittle XPath or accessibility IDs, allowing tests to adapt to UI changes. By generating tests directly from code context, the tool reduces token usage compared to earlier one‑off approaches. The open‑source project, hosted on GitHub, already powers an AI‑driven post‑development hand‑off demo that instantly verifies new features.

Developers can tie FinalRun into CI pipelines by specifying models like google/gemini‑3‑flash‑preview in config.yaml. The installer automatically sets up Android and iOS tooling, while reports are served from ~/.finalrun/workspaces. This seamless integration makes the framework a practical choice for teams seeking reliable, selector‑free mobile testing.