Product
An AI agent that tests your app the way your users use it
Green Phone Agent combines a vision-capable AI model with OneGreen's real device lab. You describe intent; the agent operates the phone, verifies outcomes and documents everything.
You
Finish the tutorial, play level 1, collect the coins and check the reward pop-up appears.
Green Phone Agent
- Launch the game on a real phone
- Skip intro, follow the tutorial hints
- Play level 1 with swipes and taps
- Collect coins — counter updates
- “Level complete” reward pop-up appears
Running on a real device
Three ways to test
Pick the mode that fits the moment
Scenario tests
Step-by-step journeys you describe in plain language, with explicit checks — ideal for critical paths like onboarding, gameplay loops and checkout.
Exploratory runs
Give a goal and a time budget. The agent explores freely and reports crashes, dead ends, broken screens and confusing UX.
Regression suites
Group scenarios into suites and run them on every build or nightly, across several device models in parallel.
Writing tests
If you can brief a tester, you can write a test
Tests are short instructions plus the result you expect. The clearer the expected result, the more useful the verdict.
Open the driver app and log in with the test account. Switch status to “Online”. Accept the next trip request and swipe “Complete trip”. Expect: the earnings screen lists the new trip and today's total increases.
Tips for great scenarios
- State the expected result explicitly
- Use test accounts and test data
- Keep one user goal per scenario
- Mention edge cases you care about (no network, app sent to background…)
It sees the screen like a person
The agent looks at each screenshot, understands buttons, text fields, dialogs and lists, and chooses the next action toward your goal. No accessibility IDs or XPath required.
- Handles permission prompts and pop-ups
- Recovers from unexpected screens
- Works with native, Flutter, React Native and web views
It proves what happened
Every run is recorded. The report lines up the video, screenshots, actions and the agent's reasoning, so developers can reproduce a failure in minutes.
- Pass / fail per check with evidence
- Shareable report links
- Reproduction steps for each bug
It runs where real users are
Runs happen on physical phones that OneGreen owns — so gestures, GPS, push notifications, GPU performance and OEM differences behave like they do for your users.
- Android and iOS devices
- Multiple models and OS versions
- Devices reset after every session
Fits your workflow
From release branch to verified build
API & CI
Trigger suites from your CI pipeline when a new build is ready and fail the job on critical regressions.
Issue tracker ready
Copy structured bug reports with steps, device info and links to video straight into your tracker.
Team workspace
Share scenarios and results across QA, developers and product owners.
Frequently asked questions
Which app types are supported?
Native Android and iOS apps, cross-platform apps (Flutter, React Native) and mobile web. Tell us about your stack during the demo.
How long does a run take?
It depends on the scenario length. A typical 5–10 step journey usually completes in a few minutes, and suites can run on multiple devices in parallel.
Can the agent make mistakes?
Yes — like any tester it can misread an ambiguous screen. That's why every verdict comes with video and screenshots, and why clear expected results in your scenarios matter.
Let an AI agent run your next test cycle
Book a 30-minute demo. Bring a build and a few test scenarios — we'll run them live on real phones.