What happened
On September 15, 2026, Claude had a bigger problem starting at 09:48 UTC. It lasted about an hour and was resolved by 11:15 UTC. Anthropic has since marked it as resolved. No detailed cause has been published beyond that.
Who this kind of outage hits, and how they usually find out
If you build automations or products on Claude, an outage like this means your workflows quietly stop producing output. No alarm goes off. Nothing in your inbox says “Claude is having trouble.” The first signal is usually a customer asking why their summary never arrived, or why the thing they paid for just did not work. That gap, between when the tool stopped and when you heard about it, is the part that costs you.
Most operators find out the wrong way. A user screenshots a blank response and sends it to you. Or you check in on something hours later and notice the queue has not moved. By then you are already behind, already explaining yourself, already wishing you had known at 09:48 instead of whenever the complaint landed.
Why this is especially rough without an engineering team
When you are not an engineer, there is no log to read. There is no error on your screen. The work just stops moving, silently, and the only way you know something is wrong is that a person tells you they are unhappy. You cannot point to a graph or a trace. You just have to apologize and say you are looking into it, while you are still figuring out what “it” even is. That is a bad position to be in, and it is the default position for most solo operators and small teams.
Timeline
- 09:48 UTC - Claude started having a bigger problem.
- About 1 hour - the disruption lasted roughly this long.
- 11:15 UTC - Claude recovered and Anthropic marked it resolved.
How a watcher catches this before your users do
NoCrash reads Claude’s official public status page every minute. The moment that page flips from working to having trouble, NoCrash sends you a plain-language message, in words you can act on, without you having to go check anything yourself. That turns “my customer told me two hours later” into “I got a calm heads-up at 09:48 and could tell people something was wrong before they had to ask.”
It also watches the things you ship. If you have n8n workflows, NoCrash watches those through your API token. If you have an app, it watches it through a URL you give it or a small JS snippet you drop in. So if the problem is on your side rather than Claude’s side, that surfaces too, in the same place, in the same plain language.
It does not find outages before the tool’s own status page does. It reads that page and tells you quickly. That is the honest version of what it does, and for most operators it is exactly enough.
The authoritative account
For the official record of this outage, go to Anthropic’s status page: https://stspg.io/x1qpycpg6ln7