> ## Documentation Index
> Fetch the complete documentation index at: https://developers.safarapi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Service Level Agreement

> SafarAPI uptime, latency and support commitments, with credits formula

## Headline commitments

| Metric                            | Target                            | Measurement                                |
| --------------------------------- | --------------------------------- | ------------------------------------------ |
| **Production API monthly uptime** | ≥ **99.5 %**                      | External probes, 3+ regions, 2-of-3 quorum |
| **p95 read latency**              | \< **500 ms**                     | Rolling 24h, server-side                   |
| **p95 write latency**             | \< **1 000 ms**                   | Rolling 24h, server-side                   |
| **Webhook delivery success**      | ≥ **99 %** within retry policy    | 7-day rolling                              |
| **Critical incident MTTA**        | ≤ **1 hour**                      | Per incident                               |
| **RTO / RPO**                     | 4 h / 15 min (5 min for WAL tier) | Per DR event                               |

The **Sandbox environment** (`sk_test_*` keys) is **excluded** from SLA — it's there for integration testing, not production traffic.

## Credits formula

If monthly production uptime falls below 99.5 %, the Partner is entitled to a credit on the next monthly invoice:

| Monthly uptime    | Credit   |
| ----------------- | -------- |
| 99.5 % — 99.0 %   | **5 %**  |
| 98.99 % — 95.00 % | **10 %** |
| \< 95.00 %        | **25 %** |

Latency breaches sustained over 4+ rolling 24-hour windows in a Service Month add:

| Sustained breach     | Credit  |
| -------------------- | ------- |
| p95 read > 500 ms    | **2 %** |
| p95 write > 1 000 ms | **5 %** |

**Maximum aggregate credit per Service Month: 25 %** of that month's settlement total.

## How uptime is measured

<Steps>
  <Step title="External probes">
    Independent probes hit `https://api.safarapi.com/api/partner/v1/health` every minute from at least 3 geographic regions (Casablanca, Frankfurt, Paris).
  </Step>

  <Step title="2-of-3 quorum">
    A minute counts as Unavailable only if at least 2 probes report failure within that minute — eliminates single-region transient noise.
  </Step>

  <Step title="Server-side latency">
    p95 latency is measured from the SafarAPI audit log (ingress → response dispatch), not from your probe location — gives you a fair view independent of your network distance to our edge.
  </Step>

  <Step title="Status page">
    Real-time dashboard and historical SLA reports on `status.safarapi.com`
    (being provisioned). Until live, SLA reports are sent to approved partners
    on request and incidents are notified directly per this SLA.
  </Step>
</Steps>

## Incident severity and response

| Severity     | Definition                                        | MTTA           |
| ------------ | ------------------------------------------------- | -------------- |
| **Critical** | Production API down, data loss, security breach   | 1 hour         |
| **High**     | One endpoint or feature unavailable, core flow OK | 4 hours        |
| **Normal**   | Degraded performance, non-blocking issue          | 1 business day |

During an incident:

1. Acknowledged on the status page within MTTA.
2. "Identified" update with root cause hypothesis.
3. Progress update every **30 minutes** until "Resolved".
4. Post-mortem within **5 business days** of resolution (Critical and High only).

## Support channels

| Channel                 | Hours                 | Use case                                     |
| ----------------------- | --------------------- | -------------------------------------------- |
| `support@safarapi.com`  | 9h — 18h CET, Mon–Fri | Integration, general questions               |
| `billing@safarapi.com`  | 9h — 18h CET, Mon–Fri | Settlement, invoicing, SLA credits           |
| `security@safarapi.com` | 24/7 ack              | Security incidents, vulnerability disclosure |
| Status page             | 24/7                  | `status.safarapi.com` (being provisioned)    |
| PagerDuty rotation      | 24/7 (Critical only)  | Triggered by external probes                 |

## Claiming a credit

Send a written request to `billing@safarapi.com` within **30 days** of month-end, with:

* the Service Month concerned;
* the measured uptime/latency figures (yours or ours);
* the calculation of the requested credit.

We respond within **10 business days** and apply the credit on the **next monthly settlement invoice**. Credits are not refundable in cash.

## Exclusions

SLA credits do **not** apply when downtime is caused by:

* Scheduled Maintenance (announced 7+ days in advance);
* Force Majeure;
* Partner-side issues (their network, their misuse, expired credentials);
* Sandbox environment;
* Beta features explicitly marked as such.

## Obtaining the SLA

The SLA is provided as a signable PDF on request — email `support@safarapi.com`.
On production approval, the countersigned copy is included automatically in your
compliance bundle.

## FAQ

<AccordionGroup>
  <Accordion title="Why 99.5 % and not 99.99 %?">
    We commit to numbers we will hit consistently while running a single-region active-passive architecture. Active-active multi-region is on the roadmap for Q1 2027 — when it ships, we'll review the SLA targets upward.
  </Accordion>

  <Accordion title="Are credits capped?">
    Yes — at 25 % of the monthly settlement total. Beyond catastrophic outages, the cap is contractual liability scope handled in the Main Agreement, not the SLA.
  </Accordion>

  <Accordion title="What counts as a write endpoint?">
    `POST /quotes`, `POST /bookings`, `POST /bookings/{number}/cancel`, `POST /webhooks`, `POST /webhooks/{id}/test`, `DELETE /webhooks/{id}`. Everything else is read.
  </Accordion>

  <Accordion title="Do you support 24/7 phone support?">
    Not in v1. Critical security incidents reach our on-call via PagerDuty 24/7. Normal/high-severity tickets are handled in business hours. We can negotiate a 24/7 commercial support tier separately if your contract requires it.
  </Accordion>

  <Accordion title="What if your status page is itself down?">
    The status page (UptimeRobot-hosted) is on independent infrastructure from the API. If both are down, that's a force-majeure-grade event — we'd reach you via email and Slack Connect directly.
  </Accordion>
</AccordionGroup>

## Related documents

* [Data Processing Agreement](/trust/dpa) — privacy and data handling.
* [Security Attestations](/trust/attestations) — TOMs detail.
* [Service status](/status) — current uptime and incidents.