Our 99.5% Uptime SLA: What We Promise, and How We Have Performed

From today, Amomic-AI has a published Service Level Agreement. The console, the public API and the AI agent runtime will be available at least 99.5% of every calendar month. If we fall below that, customers on paid plans get a credit on that month’s subscription fee.

When your agent answers the phone, replies on WhatsApp or books appointments overnight, an outage costs you real enquiries. It is fair to ask a vendor to put a number on reliability, in writing, and to pay when they miss it. That is what an SLA is for.

• 99.5%monthly uptime we commit to
• 99.994%of production requests served successfully
• 0full outages in our latest window
• ≈3h 36mdowntime 99.5% would allow per month

A promise is easier to trust when you can see the record behind it. In our latest measurement window, 99.994% of requests to our production cloud services were served successfully. Our core messaging service had a single failed request in that period, and there was no stretch in which the platform stopped serving traffic.

That number comes from the request logs of our production services, where any server error counts against us. It is a request success rate rather than a per-minute uptime figure, and a past result, not part of the guarantee. We report it because the gap between the two numbers matters to you: 99.5% is the floor we are willing to pay for, and the record shows how far above that floor we run.

We could have printed 99.99%. We chose a number we are confident we can back with money in every month, including the bad ones.

Our platform runs on Google Cloud, and every service under it carries its own published SLA. Google’s commitment for Cloud Run, for example, is 99.95%. A platform built on several services is only as available as the chain they form together, so a written promise of four or five nines on top of that chain would be marketing rather than a commitment. 99.5% is a number we can stand behind with money in a bad month, and our measured record sits far above it.

Google’s Site Reliability Engineering guidance makes the same point from the other side: chasing an extra nine costs a lot and is rarely what users notice. What they notice is whether the vendor is honest about failure and puts it right.

The rule behind the table is simple: we commit to what we operate. If WhatsApp itself is down, no vendor can deliver your messages, and an SLA that claimed otherwise would not hold up. What we do promise is that when a conversation reaches us, the agent is there to handle it.

To claim, email support@amomic.in within 30 days of the end of the month. We reply with our measured figure and a decision within 15 business days, and the service credit is applied to your future subscription invoices. Service credits are separate from prepaid usage credits. The Free plan carries the same uptime commitment, but with no fee there is nothing to credit.

Whichever AI support platform you choose, these five questions tell you whether its uptime promise means anything:

1. Is the SLA published? A number that only appears in a sales call is not a commitment.
2. Does it cover the agent, or only the website? A dashboard that loads while the agent is down is no help to your customers.
3. What is the remedy? Without credits, an SLA is only a target.
4. How is downtime measured? Per minute and per month is standard. Ask what counts as down.
5. What is excluded? Exclusions for carriers and maintenance are normal. Exclusions for “any reason at our discretion” are not.