Recorded test evidence
Inspect the recorded web and Android test results behind the agent-qa demo.
These are recorded, sanitized agent-qa demonstrations against real interfaces. They show the journeys exercised and the evidence the runtime produces. Open a test to inspect its steps, screenshots, and available artifacts in the run inspector.
This is a curated set of passing runs, not a benchmark, reliability estimate, or claim of endorsement by the products tested. It does not measure false passes or compare agent-qa with another tool. Follow the evaluation guide to test correctness and cost on your own application.
The tables count step records in each saved test. Duration is the test record's execution duration, rounded to milliseconds; it is not total setup time, suite wall-clock time, or a prediction for another run. Dates are UTC. Values reflect the saved recordings and are not live checks of these products.
Supabase
Platform: web. Recorded: 2026-05-10.
| Recorded test | Verdict | Passed steps / total | Recorded duration |
|---|---|---|---|
| Supabase login | passed | 5 / 5 | 98.745 s |
| Create Supabase Hacker News table | passed | 9 / 9 | 190.314 s |
| Insert and verify Hacker News story row | passed | 6 / 6 | 83.283 s |
Linear Android
Platform: android. Recorded: 2026-05-10.
| Recorded test | Verdict | Passed steps / total | Recorded duration |
|---|---|---|---|
| Linear Android create urgent issue and complete it | passed | 15 / 15 | 184.758 s |
Polymarket
Platform: web. Recorded: 2026-05-11.
| Recorded test | Verdict | Passed steps / total | Recorded duration |
|---|---|---|---|
| Polymarket quote prices and win amounts match live API values | passed | 13 / 13 | 257.861 s |
PostHog
Platform: web. Recorded: 2026-05-10.
| Recorded test | Verdict | Passed steps / total | Recorded duration |
|---|---|---|---|
| PostHog login | passed | 7 / 7 | 90.280 s |
| Create and verify PostHog feature flag | passed | 7 / 7 | 140.792 s |
| Capture and verify PostHog demo event | passed | 2 / 2 | 33.312 s |
Clerk
Platform: web. Recorded: 2026-05-10.
| Recorded test | Verdict | Passed steps / total | Recorded duration |
|---|---|---|---|
| Clerk login | passed | 5 / 5 | 92.933 s |
| Create Clerk user and verify through API | passed | 12 / 12 | 115.698 s |
| Add Clerk organization access and verify in dashboard | passed | 7 / 7 | 68.810 s |
Hacker News
Platform: web. Recorded: 2026-05-11.
| Recorded test | Verdict | Passed steps / total | Recorded duration |
|---|---|---|---|
| Hacker News public top story check | passed | 9 / 9 | 104.014 s |
Try it on your own app
Use the agent quickstart and its complete setup prompt to have a coding agent install agent-qa, write a test tied to a real requirement, and report the observed outcome. Keep a known failing case in your evaluation: a convincing pass needs evidence that the test can detect a regression too.