Service status
Availability is measured from outside this server, so an outage is witnessed by something that stays up when we do not. Every probe run is public.
Probed every few minutes by two independent observers since 2026-06-12, across 23,171 recorded observations.
- 24 hours · 100% of 408 probes
- 7 days · 99.960% of 2518 probes
- 30 days · 99.74% of 4537 probes
- 90 days · 99.76% of 5088 probes
Components
Each bar is one UTC day; hover for that day's probe count. Grey means no observation was made that day, shown as absence rather than counted as uptime.
The paid API answering requests: /health reachable and the catalog mounted.
Every tool route mounted and advertised on /api/pricing.
A real end-to-end purchase: challenge, payment, unlock, payload.
The hosted /mcp endpoint agents connect through.
Paid tools still answer 402 when unpaid, so nothing is given away by accident.
The chains advertised in a live 402 challenge.
A daily purchase paying real USDC across every supported chain.
Incident history
Computed from failed probes rather than written by hand, so nothing can be quietly left out. Consecutive failures are grouped into a single incident.
| Started (UTC) | Duration | Failed probes | What failed |
|---|---|---|---|
| 2026-08-02 15:02 | under 1 min | 1 | the health probe timed out - no answer before the probe's deadline |
| 2026-07-31 20:02 | under 1 min | 1 | the health probe timed out - no answer before the probe's deadline |
| 2026-07-29 03:40 | under 1 min | 1 | the /health endpoint failed |
| 2026-07-29 00:06 | under 1 min | 1 | the health probe timed out - no answer before the probe's deadline |
| 2026-07-28 22:21 | under 1 min | 1 | the health probe timed out - no answer before the probe's deadline |
| 2026-07-28 18:17 | under 1 min | 1 | the /health endpoint failed |
| 2026-07-28 16:46 | under 1 min | 1 | the health probe timed out - no answer before the probe's deadline |
| 2026-07-28 03:16 | under 1 min | 1 | the health probe timed out - no answer before the probe's deadline |
| 2026-07-28 00:22 | under 1 min | 1 | the /health endpoint failed |
| 2026-07-27 23:26 | under 1 min | 1 | the health probe timed out - no answer before the probe's deadline |
| 2026-07-27 16:51 | under 1 min | 1 | the health probe timed out - no answer before the probe's deadline |
| 2026-07-27 13:31 | under 1 min | 1 | the health probe timed out - no answer before the probe's deadline |
Live checks
Read from the running service as this page loaded, so unlike the history above these are self-reported. They are here because they lead outages rather than follow them.
Bucketed upstream balance. The amount is never exposed.
Funds paid upstream calls made on a buyer's behalf.
Funds Algorand-side external routing.
Paid with USDC.
How this is measured
- The observer sits outside the service. A GitHub Actions workflow probes production every 15 minutes and records what it saw; this page only stores and renders those observations.
- An outage appears as a gap. When production is down the probe cannot report in either, so the record shows missing observations rather than a tidy row of failures. Gaps are never counted as uptime.
- Percentages carry their denominator. 100% of three probes is a weaker claim than 100% of three thousand, and the page shows which one you are reading.
- Stale means unknown, not healthy. A component whose most recent observation has aged out is reported as not measured.
- Payment is proven with real money. A daily canary buys tools across every supported chain and settles genuine USDC. Runs are public.
- Incidents are derived, not authored. The table above is computed from failed probes, so a bad day cannot be edited out.