Use AI agents for exploratory Android testing, not as a test oracle
Use AI agents for Android exploratory testing with bounded goals, state checks, evidence capture, deterministic assertions, and honest device-farm limits.
AI agents are useful for exploratory mobile testing, natural-language journeys, and recovery from small UI changes. Keep pass or fail criteria deterministic, capture screenshots and traces, control test data, and retain conventional tests for critical assertions and regressions.
Give the agent a bounded test mission
Specify the device state, app build, account, test data, journey, expected checkpoints, and forbidden actions. Let the agent explore navigation between checkpoints, but do not let it invent what counts as a pass.
Separate exploration from assertion
A model can describe what it sees and adapt to layout changes. Critical assertions should still compare explicit text, state, network result, database state, or another deterministic signal. Store screenshots and action traces so a failure can be reproduced.
- Resettable test account and data
- Known application and OS versions
- Deterministic checkpoint assertions
- Evidence for every reported defect
- Bounded recovery and retry policy
Know what the product is not
Melaya can drive a paired real Android device and is useful for agent-assisted exploratory flows. It is not presented as an iOS solution or a replacement for a global device farm with thousands of hardware and OS combinations. Use the right infrastructure for the coverage target.
Frequently asked questions
Can AI replace mobile test scripts?
It can reduce scripting for exploratory and changing journeys, but critical regressions still benefit from deterministic tests and assertions.
What evidence should an AI mobile test retain?
Keep the app and device versions, inputs, screen captures, actions, checkpoints, errors, and the final deterministic assertion result.
Does Melaya test iOS apps?
No iOS device-control capability is claimed. Current Device Control is focused on Android.
