Read status, uptime, and run evidence
Journeys separates a verdict about your product from a problem that prevented a fair check. Down means the observed flow failed. Could not check means Journeys has no usable verdict and must not turn that uncertainty into downtime.
Check states
| State | Meaning |
|---|---|
| Up | The latest relevant scheduled check passed. |
| Down | The latest relevant scheduled check reached a failing verdict about the flow. |
| Could not check | The scheduled attempt produced no fair product verdict, for example because the browser run errored or navigation was blocked. |
| Needs account | A required tested-app credential or secret value is missing. |
| Paused | Future scheduled attempts are paused. |
| Follows another check | The check has no schedule of its own and waits for a parent check. |
| Out of credits | Journeys refused the run because the account could not fund it. |
| Key not working | The stored OpenAI key was rejected or has no usable quota. |
| Draft | The check has not completed and received approval for its learning run. |
An unknown or not-yet-judged check gets no invented status badge. On the sites list, Journeys combines real check verdicts for a site. A failing product verdict can make the site Down; a Journeys-side inability to check is shown separately and does not become a false outage.
Health strip and uptime
The health strip shows the newest 20 scheduled firings:
- Up for a passing verdict.
- Down for a failing product verdict.
- Not checked, or an unpainted gap, when no fair verdict was possible.
- Pending while the scheduled attempt is still running.
Select a painted result to open the run that produced it. Manual retries do not appear in the strip.
The check page reports rolling 24-hour, 7-day, and 30-day uptime. Only Up and Down samples are in the denominator. Journeys-caused errors, stopped runs, skipped dependencies, expired sessions, navigation refusals, credit problems, and key problems are not checked, not downtime. If fewer than half of the expected samples in a window are real verdicts, Journeys shows a dash instead of a misleading percentage.
Site History
Open History from the site to see its saved runs. Filter by check, status, trigger, or time. Status filters distinguish passed, failed or errored, still running, stopped, and not run; trigger filters distinguish scheduled, manual, and CI attempts.
An ad-hoc run of an unsaved definition has a permanent run URL, but it is not linked to a saved check and therefore does not appear in the site's History.
What a run page contains
One run page follows the attempt from queue to completion. It can include:
- The live browser while the run is active.
- Activity, including browser actions, the agent's observations, and any guidance a person supplied.
- Run facts: verdict, step count, duration, creation time, trigger, start URL, whether memory was used, variables with secrets redacted, and instructions.
- Diagnostics grouped by step: failed network requests, console warnings or errors, and uncaught page exceptions.
- A session recording and downloadable report when those artifacts were produced and are still retained.
Diagnostics are browser evidence, not automatic verdicts. A page can log a console warning and still meet the check, or fail without a console error.
After completion, the recording replaces the live view when available. Select a timed Activity or diagnostic item to seek to the corresponding page change. The recording follows browser changes and can be shorter than the run's wall clock time; older runs may not support seeking.
Detailed artifacts are retained for a limited period and may later be pruned. The run row and verdict remain, and the run page states when its artifacts were removed.