On June 23, 2026, Claude had a major outage starting at 14:19 UTC. It lasted roughly one hour and was resolved by 16:05 UTC. Anthropic has since reported the issue as resolved.
Who this hits and how they usually find out
If you build automations, workflows, or any customer-facing feature on top of Claude, an outage like this means your work quietly stops. No alarm goes off. Nothing in your interface says “Claude is down.” The first signal is usually a customer writing in to say something is broken, or you noticing hours later that a queue of jobs never ran. That gap, between when the tool stopped and when you found out, is where the damage happens.
Why it is especially rough without an engineering team
A solo operator or small team has no logs to check, no error screen to read. The work just stops moving. A prompt goes unanswered, a workflow stalls, an automated reply never sends. You do not know if it is your code, your config, or the tool itself. By the time you figure out it was Claude, you have already spent time debugging the wrong thing, and a customer has already had a bad experience. That is the shape of a quiet outage: invisible on your end, visible to your users.
Timeline
- 14:19 UTC – Claude stopped working. The outage was major.
- Roughly one hour – the disruption continued with no recovery.
- 16:05 UTC – Claude recovered. Anthropic confirmed the issue 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 immediately, sitting next to everything else you build on. You do not have to check a status page yourself or wait for a customer to tell you.
That is the limit of what it does for a tool outage, and it is an honest one. NoCrash catches the trouble within a minute of Anthropic’s own public report. It does not find the outage before Anthropic does.
On top of that, NoCrash also watches the things you ship. If you run n8n workflows, it watches those directly. If you have an app, it watches it through a URL you give it or a small JS snippet you drop in. So if something goes quiet on your own side, that surfaces too, separately from what the tool’s status page says.
The combination means “my customer told me” becomes “I got a calm heads-up first, with enough time to post a note or pause a campaign before anyone complained.”
For the authoritative account of this outage, see Anthropic’s official status page: https://stspg.io/kx1ygc7qfvx1