We do not publish an uptime figure, because we do not yet measure one. There is no monitoring agent recording availability, so any percentage here would be decoration. The live check above is the real signal: it runs when you load this page. Incident history below is maintained by hand.
We aim for high best-effort availability during early access — not a contractual SLA until redundant infrastructure and metered billing are live. See Terms §5 for details.
2026-08-05 → 2026-08-12 · scoring API unavailable. The hosting instance running the engine was terminated after the credits funding it expired. Scans returned errors for most of this window; the site and documentation stayed up. Restored 2026-08-12 on interim infrastructure. No customer data was lost — the model and code are versioned outside the host.
Current posture. Since 2026-08-14 the engine runs on Heroku (EU region, always-on dyno) — the interim founder-operated bridge is retired. Single-instance deploy: a platform restart can still cause a brief interruption, so automated pipelines should retry on 5xx.
Earlier period (2026-05-27 → 2026-08-05): no incidents recorded, but note that recording was manual and no monitoring was in place, so absence of an entry is not evidence of absence of downtime.