Status · self-checked every 30 minutes
99.24% service availability over the last 30 days.
Checks cover site pages, real-browser rendering and APIs; the probe scope, how figures are counted and when measurement began are set out below.
99.24%
1052ms
pages 1 · render 1 · API 2
Availability counts only the three checks that decide whether the service works: pages, real-browser rendering and APIs. Operational reminders from the same self-checks (a data source short, a bill coming due, a feature not switched on) do not affect whether the service is up and are not counted.
Latency
Server-side processing time, excluding network round-trip. Read the P95, not the mean — a mean is flattened by many fast requests and hides the slow tail. Timing recorded since 2026-09-10.
| Endpoint | Samples | P50 | P95 |
|---|---|---|---|
| POST /api/mcp | 8,529 | 1ms | 3ms |
| GET /api/risk | 3,193 | 51ms | 1052ms |
| GET /api/rap/conformance | 1,480 | 0ms | 1ms |
| GET /api/registry/agent | 1,398 | 2ms | 5ms |
| POST /api/decide | 263 | 36ms | 339ms |
| GET /api/x402 | 94 | 0ms | 1ms |
| GET /api/intent/check | 73 | 0ms | 1ms |
| GET /api/x402/verify-requirements | 68 | 0ms | 1ms |
| GET /api/endpoint | 67 | 74ms | 214ms |
| POST /api/intent/check | 48 | 4ms | 11ms |
| POST /api/x402/verify-requirements | 45 | 33ms | 459ms |
| GET /api/evidence/bundle | 30 | 28ms | 985ms |
| GET /api/payee/decide | 14 | 0ms | 1ms |
| POST /api/probe/plan | 13 | 109ms | 1014ms |
| GET /api/registry/agents | 11 | 6ms | 47ms |
| GET /api/baseline | 9 | 1ms | 1ms |
| GET /api/baseline/lookup | 7 | 6ms | 8ms |
What the probe covers
- 14 public pages: status code, error-page text, and whether the key content is actually present — not just a 200
- 9 endpoints: real requests, response shape checked
- 4 pages rendered in a real headless browser — this catches the “HTML fine, hydration dies” class that an HTTP probe never sees
- database, scheduled jobs, backups, risk-list freshness, mainnet lock state
- every 30 minutes on schedule, plus a silent probe every 5 minutes that only alerts on a change
Measurement scope
- Network round-trip. The figures above are server-side processing time; what you see will be slower, by an amount that depends on where you are.
- Third-party upstreams (RPC nodes, explorers). When they fail we degrade and say so, but that is not counted in the availability figure here.
- Recording began: scheduled checks 2026-09-06, latency 2026-09-10; earlier periods have no data.
- No external SLA is offered at present.