Incidents Beta

Public, rate-limited per IP

No auth required. Held to the shared public budget (600 requests/min per IP, 429 over). Submitting crowd reports is metered more tightly per IP and device to curb abuse. See rate limits.

The outage feed for a target: what providers have officially acknowledged, what our probes detected, and what other users are reporting right now.

• Base path: https://ws.latency.app/v1/incidents
• Auth: none (public)

List incidents

# Everything currently active
curl "https://ws.latency.app/v1/incidents"

Parameter	Meaning
target	Optional domain to filter to. Omit for the full active feed.

Single-target: use /details

GET /v1/incidents?target= still works, but for one host prefer the normalized details endpoint below. It returns a single incident object (not a one-element array) with the provider status page alongside.

One target (details)

curl "https://ws.latency.app/v1/incidents/details?target=cloudflare.com"

Returns everything about a single host's status in one call: the reconciled incident, the provider's own status page, its 30-day history, how stale that observation is, and a recent probe measurement.

{
  "target": "cloudflare.com",
  "incident": {
    "domain": "cloudflare.com",
    "name": "Cloudflare",
    "verdict": { "label": "Operational", "tone": "operational" },
    "reportCount": 0,
    "reporterCount": 0,
    "events": []
  },
  "statusPage": { "status": { "state": "operational", "summary": "All systems operational." } },
  "statusHistory": [ { "day": "2026-08-01", "state": "operational" } ],
  "staleness": { "stale": false, "note": "updated 40s ago" },
  "probe": { "avgUs": 41200, "samples": 20 }
}

Field	Meaning
incident	The reconciled verdict, crowd counts, and provider status-page events.
incident.events	The provider's status-page events for this host: ongoing first, then recently resolved. (Formerly pageIncidents.)
statusPage	The provider's own status page, when the host is a tracked provider.
statusHistory	Per-day page status over the last 30 days, oldest first.
staleness	How old the newest provider observation is.
probe	Recent average measurement (µs) of our endpoint probe, with the sample count.

Confirmation levels

Not every report is equal. An incident carries a status that says how sure we are:

Status	Meaning
unconfirmed	Some user reports, but below the threshold to call it real.
active	Enough independent reports to treat it as a genuine incident.
confirmed	Corroborated by the provider's own status page or our probes. The strongest signal.

The confirmed tier is what keeps the feed honest: a handful of reports won't turn a service red, and a provider quietly acknowledging an outage will surface even with zero user reports.

Reporting an outage

Anyone can contribute a report. Reports feed the crowd signal above.

curl -X POST "https://ws.latency.app/v1/reports" \
  -H 'content-type: application/json' \
  -d '{ "domain": "example.com", "type": "website", "message": "timing out" }'

Field	Meaning
domain	The affected domain. Required.
type	Reason slug, e.g. website, login, app, service, other, or a slug from /reports/recommend.
message	Optional free-text detail (most useful with type: "other").

To fetch the current suggested reasons for a target (used by our app's report sheet):

curl "https://ws.latency.app/v1/reports/recommend?query=example.com"

Crowd vs. official

The crowdVerdict you see in inspect reflects user reports only; it's deliberately kept separate from the provider's official status so a page-confirmed outage with no reports never reads as "the crowd confirmed it," and vice versa.

Related

• Catalog: the official-service status catalog of reconciliation
• Inspect: incidents in the context of a single host