# Service levels

The API is in beta. These are the targets we run it against and measure continuously. They are not a contractual guarantee: there are no service credits.

| What | Target | How we measure it |
|---|---|---|
| Availability of `/v1` | 99.5 % per calendar month (UTC) | A probe on our server calls `GET https://api.creator-guard.com/v1/` every minute, through DNS, TLS and the proxy like your requests. A minute counts as available when the answer is `200` within 5 seconds. The request log adds the share of all live requests answered without a server error (`5xx`). |
| Single-resource reads | p95 under 500 ms | Server time of every `GET` on one resource, for example `GET /v1/leads/{id}`, from the request log: 95 % of these requests finish faster. |
| Webhook delivery start | p95 under 60 s | Time from the event to the first delivery attempt, over all deliveries; manual replays are not counted. |

## What counts

- **Live mode only.** Test-mode requests (keys `cg_test_…`) run against the sandbox and are not covered.
- **Your errors are not our downtime.** Answers like `400`, `404` or `429` are the API working as documented; only server errors (`5xx`) and probe failures count against availability.
- **Webhooks:** we measure when the first delivery attempt starts, not how fast your endpoint answers. Retries after a failed attempt follow the [webhook schedule](https://docs.creator-guard.com/webhooks/).

## Status page

A public status page follows when the first partner contracts a service level. Until then, ask us about an incident at info@creator-guard.com.
