Service Level Agreement

This SLA forms part of the Terms of Service. It applies to customers on the Scale and Company plans. Starter and Growth plans receive best-effort availability without contractual credits — in practice we operate all plans on the same infrastructure, so Starter and Growth customers benefit from the same operational discipline.

Uptime is measured as (Total minutes in the month − Downtime minutes) ÷ Total minutes × 100.

A minute is counted as downtime if, during that minute, more than 5% of production HTTP health checks against the following endpoints return a 5xx status code or time out:

• GET /api/health
• GET /api/auth/providers
• GET /
• GET /skills

Health checks are performed every 60 seconds from at least 2 external monitoring locations. A minute is also counted as downtime if the skill-execution pipeline has a queue age exceeding 10 minutes for production customer workloads.

The following do not count as downtime:

• Scheduled maintenance — announced at least 72 hours in advance, performed within declared maintenance windows (Sundays 09:00–11:00 UTC unless otherwise announced). We cap scheduled maintenance at 4 hours per month.
• Customer-caused outages — misconfiguration, malformed requests, credential revocation, quota exhaustion.
• Third-party outages — AWS regional outage, Google OAuth outage, carrier-level Twilio outage. We measure and report these separately; credits may be offered at our discretion on material third-party incidents.
• Force majeure — natural disasters, war, acts of government, internet backbone failures outside our control.
• Pre-paid test / beta features explicitly labeled as such.
• Abuse — downtime caused by a customer's violation of the Acceptable Use Policy or by denial-of-service attacks targeting that customer.

If we miss the committed uptime in a calendar month, you're entitled to service credits applied to your next invoice. Credits are the exclusive remedy for downtime under this SLA.

Credits do not exceed the monthly fee for that month. Credits are applied to the next invoice and are not refundable as cash except on account closure.

Times are first-response targets, measured in business hours (Mon–Fri 9am–6pm US Mountain Time) except for S1 on Scale + Company, which is 24×7. Scale customers get a Slack Connect channel; Company customers get a dedicated CSM and private Slack.

We maintain a public status page at status.swarmengines.com showing the current health of each component and a history of past incidents. Incidents affecting one or more customers are posted within 10 minutes of detection and updated at least every 30 minutes until resolved.

To claim an SLA credit, email sla@swarmengines.com within 30 days of the end of the calendar month in question. Include:

• Your account email
• The month being claimed
• Brief description (our logs are the authoritative measurement; you don't need to submit them)

We review within 7 business days and apply the credit to your next invoice. Disputes about measurement are escalated to our engineering director for final decision.