Core concepts
Test case
A named sequence of steps with an expected result. Every test case needs an overall expected result — the acceptance criterion the AI evaluates at the end of the run to decide pass or fail. Without one, nothing is actually being asserted.
Step
One instruction. A step may optionally carry its own expected result, which is checked immediately after the step runs. Use those as checkpoints on the steps that matter; leave them blank where a step needs no verification.
Suite
A saved collection of test cases you run together — a regression pack, a smoke set. Suites can also be triggered from a CI/CD pipeline.
Run (execution)
One invocation of one or more test cases against a target URL. Every run keeps its own screenshots, verdicts and AI decision record.
Verdict
Each test case ends in one of three states:
| Verdict | Meaning |
|---|---|
| PASS | Every step ran and the expected result was met. |
| FAIL | A step failed, or the expected result was not met. |
| ERROR | The run could not complete — a timeout, or the application was unreachable. Distinct from FAIL: nothing was proven either way. |
AI engagements
Each decision the engine asks the AI to make consumes one engagement. Your remaining balance is always visible in the header. It turns amber when you are near the limit and flashes red when you are out. Runs will fail once the balance is exhausted.