Status

All systems operational

Component probes run inside the platform on every load (15s cache) — the database and queue checks are real round-trips, not a hand-set banner.

  • API● operational responding · 0ms
  • Database● operational responding · 11ms
  • Queues & realtime (Redis)● operational responding · 1ms
  • Web (console · buyer · Contract Link)● operational you are reading this page from it
  • Escrow custody● operational non-custodial — funds at regulated partners, never on Pontis infrastructure

Measured, not promised

API p95 latency · 24h

467 ms

22 calls · p50 156 ms

API error rate · 24h

0.00%

5xx share of recorded calls

Webhook delivery · 7d

7.61%

14'084 deliveries attempted

6'587 events hash-chained · 1'114 releases on proof · 71 audit anchors · last anchored Fri, 17 Jul 2026 00:17:08 GMT (across live and sandbox deals)

Incident history

Written by hand, kept honest. Post-mortems name the cause and the fix — and what changed so it can't repeat.

  • 2026-07-10 · ~2 hours from first dispatch to guard deployed

    Test-queued emails dispatched by the production notify worker

    Impact: A handful of notification emails generated by automated tests reached real inboxes. No customer data was exposed and no money movement was affected.

    Cause: The test suite and the production worker shared one database; the worker drained QUEUED notification rows without distinguishing their origin.

    Fix: A suppression guard shipped the same day (test-origin recipients are never dispatched), and CI moved to ephemeral copy-on-write database branches so tests no longer share the production data plane.