Skip to content
OpenKey

Status

A plain, indexable page — no widget, no third-party subdomain. What's here is real: current state now, uptime history as it accrues, incidents when they happen.

All systems operational

MONITORING SINCE JULY 2026

Endpoints

Each endpoint gets one bar per day across a 90-day window. This page launched in July 2026, so the history is young — bars fill in as days pass. No percentage appears until it's measured.

  • api.openkey.ai/v1/chat/completionsInference API

    Operational

    90-DAY UPTIME — COLLECTING SINCE JUL 2026

  • api.openkey.ai/v1/modelsModel catalog API

    Operational

    90-DAY UPTIME — COLLECTING SINCE JUL 2026

  • dashboard.openkey.aiDashboard

    Operational

    90-DAY UPTIME — COLLECTING SINCE JUL 2026

  • openkey.aiThis site

    Operational

    90-DAY UPTIME — COLLECTING SINCE JUL 2026

Incidents

No incidents recorded yet.

When they happen, they'll be listed here with real causes — named providers, named failures, timestamps — not "degraded performance" euphemisms. That's part of ourerror-semantics contract.

What we commit to

From the error-semantics contract — a contract, not documentation of best effort.

  1. 5xx means us, 4xx means you — always. An infrastructure failure never masquerades as an auth error.

  2. Rate limits fail loud. A 429 with Retry-After — never silent queueing, degrading, or rerouting.

  3. No silent model substitution. Fallback routing is opt-in per request. If a provider is down, the error names it.

  4. Every error carries a machine-readable cause. Code, provider, request id, and whether a retry is worth it.

  5. This page updates first during incidents. Post-mortems name the real cause, and contract violations are owned in the changelog.