Service level agreement.
What uptime we promise, and what you receive if we miss it. Effective 1 September 2026.
1. What this agreement covers
This agreement sets out the monthly uptime we promise for cloud servers and for the network. It forms part of the terms of service and applies to every account. Uptime is measured for each region, over each calendar month. The 6 regions are in Tallinn, Riga, Copenhagen, Vienna, Milan and Lisbon.
2. What we promise
| Service | Monthly uptime |
|---|---|
| Cloud servers | 99.95% |
| Network | 99.99% |
A server is unavailable when it can’t be reached on any of its addresses because of a fault on our side. Each minute counts once our monitoring has confirmed the fault.
The network is unavailable when a region can’t send or receive traffic to the internet or to the other regions. Our monitoring checks every minute from three locations outside the platform. The results are on the status page, and GET /v1/status/incidents returns each incident with its region and its minutes.
3. Service credits
If we miss a promise, you receive a credit on the month’s charge for the resources it affects.
| Monthly uptime of a server | Monthly uptime of the network | Credit |
|---|---|---|
| Below 99.95% | Below 99.99% | 10% |
| Below 99.5% | Below 99.9% | 25% |
| Below 99.0% | Below 99.5% | 50% |
For a server, the credit applies to that server’s charge for the month. For the network, it applies to the charge for every resource in the affected region.
The credits in a month never add up to more than half of that month’s charge for the affected resources. Credits go on your next invoice; they are not paid out in cash.
4. What does not count
These events don’t count against the promise:
- maintenance announced on the status page at least five days ahead;
- a fault in your own software, configuration or firewall rules;
- an attack you cause, or a suspension under the terms of service;
- an event outside our reasonable control, such as a regional power failure;
- features we mark as a preview in the console.
During maintenance, we move running servers to another host where we can, so most maintenance causes no downtime.
5. How to claim a credit
Open a support ticket within 30 days of the end of the month. Name the resources, the region and the times they could not be reached. We check the claim against our monitoring and reply within ten working days. If the claim holds, the credit appears on your next invoice.
6. Other services
Storage, databases, Kubernetes and load balancers run on the same servers and network. They have no separate promise in this agreement, but their incidents appear on the status page all the same.