Service Level Agreement

Last updated 2026-06-12 · applies to cogos.5ceos.com · self-serve tiers

This page describes the uptime objectives CircaOS commits to on each self-serve tier, how we measure them, and what remedies you get if we miss them. Enterprise contracts may carry separate or stricter SLAs in the executed order form — those override anything here.

What "uptime" means here. A 5-minute interval counts as DOWN if more than 50% of authenticated /v1/* requests against cogos.5ceos.com from healthy clients fail with HTTP 5xx (excluding rate-limit responses, customer-side authentication failures, and scheduled maintenance). Monthly uptime is the percentage of 5-minute intervals in the calendar month that are UP.

Per-tier SLOs

Tier	Monthly uptime SLO	Support response	Remedy when missed
Developer $29 / mo	99.0% ~7 h/mo budget	Community (forum, email best-effort)	Hard cap on overage charges; pro-rated credit on request.
Team $199 / mo	99.5% ~3.6 h/mo budget	Email support, next-business-day response	10% monthly credit if SLO missed; 25% if <99.0%.
Business $999 / mo	99.9% ~43 min/mo budget	Priority email, same-business-day response	15% monthly credit if SLO missed; 35% if <99.5%; 75% if <98.0%.
Enterprise contract	99.95%+ per contract	Per executed order form (named technical sponsor available on request).	Per executed order form — may include financial penalties beyond service credits.

How we measure

Probe. A Container App Job runs continuous external probes against cogos.5ceos.com from a separate Azure region. Each probe issues an authenticated GET /v1/me and records the HTTP status, latency, and any error envelope. Probe cadence: every 60 seconds.
Real-traffic floor. Probe data is the floor of the measurement, not the ceiling — if real customer traffic shows higher error rate than probes detect, we use the customer-reported figure for credit calculation, not the lower probe number.
Receipt-bearing. Every /v1/* response carries an Ed25519 attestation receipt (header X-Cogos-Attestation) and an HMAC tamper signature (header X-Cogos-Signature). Both bind the response to the running build's chain head. If a customer claims downtime and we have no audit-chain rows for that window from their tenant, we credit on the customer's report; if we DO have rows, the cryptographic receipt is the source of truth for whether the service was up for THAT customer.

Scheduled maintenance

We may take the service down for maintenance with at least 48 hours' email notice. Scheduled maintenance does NOT count against the SLO. Today, planned-maintenance windows are rare — the runtime is container-rolled with zero-downtime revisions whenever possible. Cap: no more than 2 hours of scheduled maintenance per calendar month without a separate notification.

Exclusions

The SLO does NOT cover the following:

Customer-side authentication failures (HTTP 401), rate-limit responses (429), or per-tier quota responses (402). These reflect the service correctly enforcing the customer's plan.
Outages caused by force majeure, DDoS in excess of edge capacity, cloud-provider regional incidents on Azure, or upstream incidents the substrate routes through that we communicate within 30 minutes of detection.
The :real live-search fallback path's external search provider (the substrate's resolution path doesn't depend on the search provider being up, but a customer who explicitly forces live-search may see degraded results if the provider is degraded).
Beta and preview endpoints (anything documented as "preview" or "experimental"). These ship without uptime commitments.

How to claim a credit

Email support@5ceos.com within 30 days of the incident with your tenant slug and the 5-minute intervals you believe were down. We respond within 5 business days with our own measurement for the same window, the calculated credit if any, and a link to the audit-row excerpt for your tenant during the window. Credits are applied to the next invoice; they are not paid out as cash.

One honest gap

We don't yet publish a real-time status page. The Trust page shows the running process's uptime and gives a snapshot of build & advisories, but a multi-incident timeline is on the roadmap and not shipping today. Customers can subscribe to our status mailing list at support@5ceos.com in the meantime.