Uptime targets, measured monthly, credited when missed.

This Service Level Agreement ("SLA") applies to the HyperliquidRPC data services (gRPC streams, HyperEVM JSON-RPC, HyperCore /info, and the historical query API) for customers on the Pro and Ultra plans, and for Enterprise contracts. Uptime targets are 99.9% on Pro and 99.99% on Ultra.

For Enterprise customers, an order form may set different targets (for example, for a dedicated node); where it does, the order form controls.

• Surface: each of the four service planes, measured independently: streams (gRPC), HyperEVM RPC, HyperCore /info, and the historical query API.
• Unavailable: a surface returns errors or no data for sixty consecutive seconds, as measured by our external monitoring. Elevated latency alone is not Unavailability; latency is tracked separately.
• Monthly Uptime Percentage: for a surface, 100% minus the percentage of minutes in the calendar month in which that surface was Unavailable, after subtracting excluded minutes under Section 5.

For each surface and each calendar month, the target on Pro is 99.9% Monthly Uptime Percentage, and on Enterprise it is 99.99%. Our monitoring is the record we settle credits against.

If a surface misses its target in a calendar month, you are entitled to a credit against that month's subscription fee: 10% if uptime was below target but at least 99.0%; 25% if below 99.0% but at least 95.0%; 50% if below 95.0%. Credits are applied to a future invoice, are capped at 50% of the affected month's fee in aggregate, and are not paid as refunds. Service credits are your sole and exclusive remedy for missed availability targets.

Unavailability caused by the following does not count:

• Scheduled maintenance announced at least 48 hours in advance via email (Section 6).
• Upstream events on the Hyperliquid network itself: chain halts, degraded consensus, or outages of Hyperliquid's public infrastructure that no provider can serve around. We surface these in incident communications as upstream events.
• Your own configuration, key suspension under the Acceptable Use Policy, or exhaustion of your tier's quotas (rate limiting at your plan's ceiling is the service working as specified).
• Force majeure and failures of networks or systems outside our control, other than our own subprocessors.

Planned maintenance is announced by email at least 48 hours ahead, scheduled in low-activity windows, and engineered to avoid interruption where possible. Emergency security patching may proceed on shorter notice; we notify customers before or as it happens.

Email support with the subject "SLA credit", the calendar month, the affected surfaces, and request timestamps, within 30 days of the end of the affected month. We reconcile against our monitoring and confirm the credit within ten business days.

We version this SLA (header above) and notify customers of material changes at least 30 days before they take effect. Changes never apply retroactively to a month already in progress. The current version always lives at this URL.