Uptime,
in writing.

What’s covered.

Each covered product has its own definition of unavailable. Each of your resources of a covered product has its own monthly uptime and its own credit.

How we measure it.

• We check every covered resource every minute. Apps are checked from two independent locations: if either sees your app serving, the minute counts as available. Databases and caches are checked by the location that speaks their own protocol, and a minute counts against them only while the servers they run on were reachable from it.
• A minute in which a resource had no result while our checks were otherwise reporting counts as unavailable. A minute in which our measurement itself delivered nothing is a measurement gap: never counted against you, and shown to you as one.
• A private database or cache, or one open only to your own IP addresses, is measured through its network path: the check confirms the path into the servers it runs on answers, not the database itself.
• Each resource’s uptime is calculated per calendar month in your account’s own time zone. A minute it was not running, or did not exist yet, counts as available.
• A resource counts from its first successful check once it is ready, until you delete or stop it.

A resource’s monthly uptime % = (minutes in the month − its unavailable minutes) ÷ minutes in the month × 100

Credits.

When a resource falls short in a month, its credit is a percentage of that resource’s own charges for the month, less anything paid with promotional or migration credits. Startup and Business are credited automatically from November 2026; an Enterprise contract sets its own terms.

Credits are capped at 50% of that resource’s charges for the month, start at US$1, and are the sole remedy for unavailability.

Paid automatically.

1. 1
   We record every minute.

   External checks run every minute, all month long.

2. 2
   We calculate after the month ends.

   Early in the next month, we work out each resource’s uptime for every workspace on Startup or Business.

3. 3
   Credit lands in your wallet.

   Your wallet statement shows the credit with each resource’s uptime and the percentage applied.

Want a second look? Open a support ticket within 30 days of the month end and we’ll review it.

Credits go straight to your Plexivo wallet and apply to any Plexivo usage.

Always running.

Your workloads run independently of the Plexivo console and API. If they are ever unavailable, what your customers use keeps working.

The Plexivo console, the management API, the Client Data API and Plexivo AI wait until the console and API are back. They are not part of the 99.9% customer-resource commitment.

What isn’t counted.

• Your own actions. From the moment you ask until the resource is healthy again: delete, stop and start, restart, redeploy or roll back, resize or scale, a configuration change that restarts it, a restore, and a version upgrade or patch you apply now. The same goes for your project or organisation being deleted.
• Scheduled maintenance announced in advance, inside its announced window, as set out below.
• Measurement gaps: minutes in which our measurement itself delivered nothing.
• Your own app. Errors your code returns, and problems caused by your code, configuration, dependencies or data.
• Management services. The Plexivo console, the management API, the Client Data API and Plexivo AI.
• The Developer plan, products not listed as covered, and features marked Beta.
• Resources paused because your wallet balance reached zero or you went over your plan’s limits.
• Events outside reasonable control, such as natural disasters or government action.

Scheduled maintenance.

• We announce maintenance by email and in the console at least 5 days ahead, and do it in your resource’s maintenance window, which you can choose.
• Only the minutes inside the announced window are not counted, and every exclusion together leaves at most 4 hours a month out of a resource. Any minute past the window counts.
• Maintenance announced with less notice counts in full, including an urgent security fix.
• Your console shows each one as “Scheduled maintenance (announced date)”, with every other exclusion that affects you. You can also apply an update earlier yourself; that is your own action.

Check our record.

Changes to this SLA.

We give 30 days’ notice before any change that reduces your coverage. Accounts that accepted the earlier version (2026-10-17) keep its terms, where they give more, until 26 November 2026. Older versions are available on request.