1. Definitions
These definitions do the real work in this document. Read "Downtime" and "Gateway Error Rate" together - they decide whether a credit is owed.
For prepaid accounts, charges actually deducted from your available balance for API usage during that month, excluding unused top-up balance, taxes, refunds and prior credits. For postpaid enterprise accounts, the charges calculated for that month under the applicable agreement, before credits.
The percentage of the calendar month during which the gateway was not in Downtime, calculated under Section 2. Total Minutes means every minute of that calendar month.
At least 5 consecutive minutes with Gateway Error Rate above 10%. Measured in whole minutes. Excludes anything under Section 4.
Failed Eligible Requests / total Eligible Requests in a one-minute interval, across the entire gateway. Not calculated per model.
Either (a) a response carrying HTTP status 500, 502, 503 or 504 that originates in RouteAPI infrastructure, or (b) no response within 30 seconds - connection refused or reset, TLS handshake failure, DNS failure, or any transport-level failure that yields no HTTP status at all. Not a Gateway Error: responses caused by request or authorisation (400, 401, 403, 422), rate limiting (429), or an error that originates upstream even though RouteAPI returns it to you with the same status code.
Well-formed request to a production endpoint, valid API key, not rejected for customer-attributable reasons.
RouteAPI synthetic probes issued through the same public ingress you use, at intervals of no more than 1 minute, each request subject to a 30-second response timeout. Published at routeapi.ai/status.
2. Service Commitment
RouteAPI will use commercially reasonable efforts to make the Service available with a Monthly Uptime Percentage of at least 99.9% in each calendar month.
(Total Minutes − Downtime Minutes) / Total Minutes × 100This commitment applies to the gateway as a whole, not to any individual model. The Service is considered available at any given time if at least one model supported by the gateway is capable of accepting and successfully processing requests.
Rationale. The value proposition of an AI gateway is redundancy and flexibility: when one model is unavailable due to upstream rate limits, regional outages or provider maintenance, you can route to alternatives without interruption. RouteAPI provides multiple upstream channels for most models, which significantly reduces the likelihood that all paths to all models fail at once. Measuring at the gateway level reflects this design and matches your actual experience — the ability to get a response.
Transparency. We publish real-time availability for each individual model on the public status page, including uptime, error rates and incident history. Use it to monitor what you depend on. But service credits are calculated solely on gateway-level availability, not on the availability of any single model.
For reference, 99.9% corresponds to a maximum of approximately 43.2 minutes of Downtime in a 30-day month and 44.6 minutes in a 31-day month.
3. Service Credits
If the Service Commitment is not met in a calendar month, you are eligible for a Service Credit calculated as a percentage of your Eligible Usage Charges for that month.
| Monthly Uptime Percentage | Service Credit |
|---|---|
| Less than 99.9% but ≥ 99.0% | 10% |
| Less than 99.0% but ≥ 95.0% | 25% |
| Less than 95.0% | 50% |
- < 99.9% · ≥ 99.0%10%
- < 99.0% · ≥ 95.0%25%
- < 95.0%50%
Applied as a credit to your account balance within [30] days of approval, usable for future use of the Service. We do not offset it against a future invoice — prepaid accounts have no invoice to offset.
Service Credits are your sole and exclusive remedy for any failure to meet the Service Commitment.
May not exceed 50% of the Eligible Usage Charges for the affected month.
Not cumulative with any other credit for the same period.
4. Exclusions
The Service Commitment does not apply to, and Downtime does not include, unavailability caused by the following. Item 4.2 is the one worth reading closely — it is the only exclusion where the burden of proof sits with us.
Scheduled Maintenance
Announced at least [48 hours] in advance via the status page and email to your designated contact, and not exceeding [4 hours] per calendar month in aggregate. Production nodes are maintained one at a time, so routine maintenance does not interrupt traffic.
Upstream Model Provider Failure
Burden of proof: RouteAPIUnavailability, rate limiting, errors or degraded output of the third-party model provider you selected — but only to the extent confirmed by that provider's publicly available status page or incident report. RouteAPI bears the burden of demonstrating such confirmation.
Customer-Attributable Causes
Invalid requests (4xx), exceeded quota or rate limits, suspended or unfunded accounts, misuse, or use in violation of the Terms of Service.
Force Majeure
Events beyond RouteAPI's reasonable control: natural disasters, war, terrorism, labor disputes, governmental action, Internet backbone or DNS failures outside our network, and denial-of-service attacks that cannot be reasonably mitigated.
Beta, Preview or Deprecated Features
Any model marked "preview", "experimental" or "deprecated" in the model listing.
Your Network, Software or Equipment
Third-party services not under RouteAPI's control.
5. Credit Request Procedure
Credits are not applied automatically. Submit a request within thirty (30) days after the end of the calendar month in which the Downtime occurred.
Email us within 30 days
Send to support@routeapi.ai within thirty (30) days after the end of the calendar month in which the Downtime occurred.
Include the evidence
Your account identifier; the dates and times of each Downtime incident claimed; and request IDs with timestamps, or server logs documenting the errors.
We review and respond
We check your request against our Measurement Source and our own request logs, and respond within [15] business days. If we confirm the Service Commitment was not met, the credit is applied to your account balance per Section 3(a).
6. Status Page and Measurement
Every number in this SLA comes from one place: RouteAPI's synthetic probes, published publicly. If our measurement and yours disagree, we say so and show ours.
The status page is provided for transparency. In the event of a discrepancy, RouteAPI's internal measurement records govern.
7. Changes to this SLA
RouteAPI may modify this SLA with at least [30] days' notice. Modifications will not reduce the Service Commitment for any Enterprise Customer during the then-current term of its Enterprise Agreement.
FAQ
The questions enterprise reviewers ask most often, answered without hedging.
The point of a gateway is that you are not dependent on one model. Our commitment is that you can always get a response — and if the model you asked for is unavailable, you can route to another. Per-model availability is published on the status page so you can monitor what you depend on, but the credit calculation follows the gateway-level definition in Section 2. This is also how AWS measures API Gateway availability: by region, not by individual backend.
We are an aggregation layer, not an owner of the underlying compute: a large share of what you pay goes to the upstream provider. Credits of 10% / 25% / 50% already take us to the edge of what the business can absorb while continuing to operate. We would rather commit to a number we can pay than publish a number we would have to argue about. Credits are also applied to your balance rather than refunded, so they remain available for you on future usage.
No. This SLA applies to customers with an executed Enterprise Agreement or Order Form that references it. Self-service and prepaid accounts, free tiers, trials and beta features are not covered. Availability for all accounts is published on the status page under the same measurement method — the difference is the contractual remedy, not the measurement.
You do not have to prove it — we do. Section 4.2 excludes an upstream provider failure only when that provider's own public status page or incident report confirms it, and RouteAPI bears the burden of demonstrating that confirmation. Anything failing inside our own aggregation, routing or supply layer counts as Downtime, full stop.
On its own, it does not. Downtime requires the Gateway Error Rate across the entire gateway to exceed 10% for at least five consecutive minutes — meaning the gateway as a whole could not return responses. A single model being unavailable is visible on the status page and you can route around it, but it is not Downtime under this SLA.
Contractual 99.9% commitment with service credits, for customers on an enterprise agreement.
Request agreement