Public API · Availability
100% / 99.9% targetmetMeasured over a rolling 30-day window.
100% of the error budget remains for this window.
Platform status · Aug 27, 2026 · 15:01 UTC
What we promised, and what actually happened.
Every figure below is computed from recorded measurement windows by the database itself. None of it is entered by hand, which means an objective we are missing shows as missed. The same data is available without a credential at /api/v1/status.
An objective is breached when measured attainment falls below its published target over the rolling window, and an incident stays open until it is resolved here.
Service level objectives
Webhook delivery · Delivery success
25% / 99.5% targetbreachedMeasured over a rolling 30-day window.
0% of the error budget remains for this window.
Compliance scanning · Scan success
— / 99.9% targetunmeasuredTarget published over a rolling 30-day window. No observations have been recorded against it yet, so there is no attainment to report.
Incident history · last 90 days
No published incident in the last 90 days.Incidents are published here with their timeline, not summarised after the fact — every update is append-only in the database that serves this page.
How these numbers are produced
Each objective states the fraction of good events required over a rolling window. Observed counts are written to measurement windows, and attainment is derived from them by a database function — it is never stored as a figure someone can edit. An objective with no windows in range reports unmeasured, which is why you will sometimes see grey here instead of green.
The error budget is the share of permitted failure still unspent. At 100% nothing has gone wrong in the window; at 0% the objective is exactly at its target and the next failure breaches it. Incident updates are append-only: once posted, an entry cannot be edited or deleted, because a timeline that can be rewritten afterwards is a press release.
Two limits worth stating plainly. API responses served from the edge cache never reach our code and so are never counted — the sample is the traffic that reached the origin, which is the traffic that could have failed, so the published figure understates availability rather than flattering it. And requests we reject for a bad credential or a malformed body are excluded entirely: those callers did not experience an outage, and counting them would make this page describe our callers instead of our service.