Service-level targets

This document records the availability targets EchoRelay operates against. It is a published target, not a contractual commitment — enterprise plans can negotiate a formal Service Level Agreement with service credits as part of their order form. For everyone else, this page is what we hold ourselves to.

Target availability

Surface	Target	Window
Data plane — *.echorelay.cloud	99.9%	Rolling calendar month
Management plane — panel, REST, MCP	99.5%	Rolling calendar month
Documentation — docs.echorelay.dev	Best-effort	—

What "available" means

The data plane is considered available when an authenticated relay request returns a response within the documented latency budget and the request is successfully enqueued for delivery. A response with HTTP 5xx attributable to EchoRelay infrastructure (not the customer-configured target) counts as downtime; a 4xx caused by caller misconfiguration or quota exhaustion does not.

The management plane is considered available when a logged-in user can read or write project configuration through the panel, the REST management API, or the MCP endpoint.

Maintenance windows

Scheduled maintenance is announced at least 48 hours in advance on the status page and counts as downtime against the management-plane target but not against the data-plane target (we aim to keep the queue draining through every planned change). Emergency security patches may be applied without notice; we publish a postmortem within 7 days.

Exclusions

The targets above exclude:

• Outages or degraded performance of customer-configured target URLs (these are surfaced as 5xx responses from your target, not from us).
• Outages of upstream third parties — Cloudflare, Paddle, Hetzner — provided we apply mitigations within a reasonable window.
• Caller-induced throttling: rate-limit denials, credit exhaustion, key revocation.
• Scheduled maintenance announced 48 hours in advance.
• Force majeure as defined in the Terms of Service.

Reporting

Availability is measured via independent third-party probes and our own internal monitoring. Incidents that affect a meaningful slice of customers are recorded on the status page within 15 minutes of detection, updated through resolution, and followed up with a written postmortem within 7 days.

Service credits

No automatic service credits are offered against this published target. If your contract or order form references a Service Level Agreement with credits, those terms govern. Enterprise customers can request the negotiable SLA template by emailing [email protected].