Loading status…
Fetching the latest health data.
Every number above is computed from real solve traffic and real health probes, not from a schedule of planned uptime. The page repolls every 30 seconds; nothing here is cached longer than that.
Four surfaces are tracked independently: the Cloudflare Turnstile solver, the Turnstile WAF solver, the REST API that accepts and returns solves, and the dashboard. A degraded solver does not imply a degraded API, so they are reported separately rather than rolled into one figure.
The median wall-clock time from POST /v1/solve to a returned token, over a rolling 24 hours. Median rather than mean, because a handful of retried solves would flatter or wreck an average without telling you what a typical request costs.
Solves that returned a usable token, over solves attempted. Failed solves are never billed, so this figure and your invoice move together.
Ninety days, one cell per day, built from stored health samples. Days before a component started reporting are shown as unknown rather than filled in as healthy — an absence of data is not an absence of incidents.
Incidents are written by the team as they are worked, and stay on the page after they resolve. For programmatic checks, the API reference documents the health endpoint; for changes to the service itself, see the changelog.
These are our probes, so they report what our probes reach and nothing else. If your solves are failing while every component above reads healthy, that gap is the interesting part — send us the solve ids rather than waiting for a status change that may never come.