# Status API

`GET /v1/status` returns the same data as the [status page](/status), as JSON. It needs no API key, costs
nothing and can be called from a browser on any site. Answers are cached for 30 seconds, so there is no
point asking more often.

```sh
curl "$API_BASE/v1/status"
```

```json
{
  "generated_at": "2026-11-16T10:00:50Z",
  "overall": "operational",
  "components": [
    {
      "id": "api",
      "name": "API",
      "state": "operational",
      "since": "2026-11-01T08:00:00Z",
      "detail_code": null,
      "uptime_90d": 99.98,
      "days": [{ "date": "2026-11-15", "state": "operational", "ok_minutes": 1440, "total_minutes": 1440 }]
    }
  ],
  "registration": { "checked_at": "2026-11-16T06:00:00Z", "tlds_total": 43, "tlds_ok": 42, "failing_tlds": ["pt"] },
  "incidents": { "open": [], "recent": [] }
}
```

## Fields

| Field | What it holds |
| --- | --- |
| `overall` | The worst state of any component. |
| `components` | One entry per part of the service: `web`, `api`, `mcp`, `jobs`, `monitoring`, `browser`, `dns`, `registration`, `email` and `payments`. |
| `components[].state` | `operational`, `degraded`, `outage`, `maintenance`, or `unknown` while there is no recent data. |
| `components[].detail_code` | A fixed code saying why a component isn't operational, such as `slow_responses`, or `null`. |
| `components[].uptime_90d` | The share of minutes the component was working over the last 90 days, in percent. |
| `components[].days` | One entry per day: `operational` at 99.9 % of minutes or more, `degraded` at 99 % or more, otherwise `outage`; `no_data` for a day without samples. |
| `registration` | The nightly registry check: how many TLDs answered as expected, and which didn't. |
| `incidents` | Open incidents and maintenance, and those of the last 14 days, each with its updates. |

## Using it

- Poll at most every 30 seconds; the answer won't change faster.
- Direct callers are limited to 120 calls a minute.
- If the API itself can't be reached, the status page shows the last snapshot it had. Treat a failed call
  to this endpoint as a sign of trouble too.
