Latency by endpoint
Median and 95th percentile, measured server-side. The cheap reads answer in single-digit milliseconds; the deep analysis endpoints fan out to dozens of upstream calls and take seconds. That gap is the real shape of the product, so it is published per endpoint rather than blended into one number that would describe neither.
| Endpoint | Requests 30d | p50 | p95 | 5xx |
|---|---|---|---|---|
| Loading live figures from api.defade.org… | ||||
What the error rate means
Share of requests answered without a 5xx. Measured from requests we served, so it cannot observe an outage in which nothing reached us — it is a server-error rate, not an uptime guarantee. Better a number that means something narrow than a percentage implying more than it can support.
What the latency means
Server-side, per endpoint, in milliseconds. Deliberately not averaged into one figure: the cheap reads answer in single-digit milliseconds while the deep analysis endpoints fan out to dozens of upstream calls, and a blended number would describe neither.
Coverage
Endpoints with fewer than 20 requests in the window are omitted — percentiles over a handful of calls say nothing. Figures come from the request log and are recomputed at most every ten minutes.
Raw JSON: api.defade.org/api/status · Full API documentation · Plans from $29/mo