System status
Every number below is measured on this deployment right now, not copied from a status template. The JSON behind it is /v1/status.
Gateway and signer
+1.52 msproxy overhead p50 · p95 3.42 ms · p99 3.68 ms (published benchmark)
1.08 mscache hit p50 · answered without a model call
4workers online · uptime 318 min
verifiedEd25519 signer self-test · key ae3jyGdg2kH6wJne
62 modelspricing catalog · 2026-10-26 · fresh · 5 providers
+1.98 msstreaming first-byte delta p50 · gateway minus direct
Upstream providers and automatic failover
Failover is opt-in per request or per workspace (x-tokenzero-fallback: auto); rescued requests are counted when a retry or a cross-provider re-issue turned a 429/5xx into a 2xx.
| Provider | Status | Models tracked | Pricing | Retries 24h | Rescued 24h | Breaker | Failover policy |
|---|---|---|---|---|---|---|---|
| OpenAI | operational | 30 | fresh | 0 | 0 | closed | retry 429/5xx with jitter; cross-provider to any OpenAI-shaped model you name |
| Anthropic | operational | 14 | fresh | 0 | 0 | closed | retry 429/529/5xx with jitter; same-shape fallback only |
| Google Gemini | operational | 10 | fresh | 0 | 0 | closed | retry 429/5xx with jitter; OpenAI-compatible and native routes |
| xAI | operational | 6 | fresh | 0 | 0 | closed | retry 429/5xx with jitter; cross-provider to OpenAI-shaped models |
| DeepSeek | operational | 2 | fresh | 0 | 0 | closed | retry 429/5xx with jitter; off-peak pricing honoured |
Self-serve diagnostics
Connection and receipt doctor
Six checks in one call: gateway, signer, pricing, your key, quota and settings.
curl -s https://api.tokenzero.sh/v1/doctor -H "X-TokenZero-Key: $TOKENZERO_KEY"
npx -y @tokenzero/cli doctor
Support
Open a ticket on /support or by API (POST /v1/support); engineering targets a first response within 15 minutes. Email support@tokenzero.sh · GitHub issues for the open-source tools.
Receipts verify offline with the public key; the chain walks with tokenzero verify --chain.