Finalrun-agent turns plain English into visual mobile tests

By vramkickedin | | 2 min read
Extreme close-up of a smartphone corner angled engraved into this surface is the text Finalrun-agent.

Finalrun-agent is an AI-driven command line tool that tests Android and iOS applications using natural language. Users write tests in plain English within YAML files, and the tool launches the app on an emulator or simulator to execute each step visually. It supports AI models like Gemini, GPT, and Claude to see the screen, perform actions like tapping and swiping, and generate pass/fail reports with video recordings.

The tool was created by developers frustrated with traditional mobile testing methods that break whenever small UI changes occur. Instead of relying on fragile CSS or XPath selectors, finalrun-agent visually understands what appears on screen and follows user intent. This approach helps catch issues that standard automation often misses, such as layout problems and misaligned elements.

Testing features that adapt to your app

  • Write tests in plain English using YAML format.
  • Support for both Android and iOS platforms.
  • Integrates with Gemini, GPT, and Claude AI models.
  • Generates video recordings and device logs with reports.
  • AI coding agent skills for automatic test generation.
  • Bring your own API key for your preferred AI provider.
  • Works with popular IDEs like Cursor and Codex.

Mobile developers and QA teams working across Android and iOS can save significant maintenance time with this approach. Small agencies managing multiple apps may find the visual testing method reduces the overhead of keeping selectors updated after each UI change.

Why the team built this

The developers created finalrun-agent after experiencing a common frustration in mobile development. Small UI changes would break existing tests, forcing teams to constantly fix selectors while other parts of the test suite continued failing.

"Instead of relying on fragile CSS/XPath selectors, we built Finalrun,"

the team explained in their announcement post.

They noted that because the agent actually sees the app visually, it can catch UI and UX issues that typical automation tends to miss. The project is now open source under the Apache license, and the team welcomes feedback from developers dealing with flaky test suites.

You can access finalrun-agent on GitHub to get started with visual mobile testing.