what happened
On the evening of June 16, 2026, Claude had a bigger problem that lasted about 46 minutes. It started at 20:50 UTC and was resolved by 21:37 UTC. Anthropic has since confirmed the issue is resolved. The status note gives no further detail on the cause, so neither will this.
who this hits and how they usually find out
If you use Claude inside a workflow, a product, or a client-facing tool, you probably did not build a way to know when Claude stops working. Most people who build on an API like this find out the same way: a customer writes in to say something is broken, or a task that should have finished hours ago is just sitting there. By the time you hear about it, the outage is already over, but the damage, a missed delivery, a confused customer, a broken promise, is already done. The tool itself rarely sends you a message saying it is having trouble.
why this is especially rough without a technical background
If you are not an engineer, there is no log file to open, no error screen to read. The work just stops moving. A Claude-powered summary does not arrive. A draft does not get written. An automation that runs every hour quietly skips a run. Nothing announces itself. The first signal is often a customer asking why something did not happen, and by then you are already behind, explaining a problem you did not know existed until thirty seconds ago. That gap, between the moment something breaks and the moment you find out, is where the real cost lives.
timeline
- 20:50 UTC, June 16, 2026: Claude starts having a bigger problem.
- 20:50 to 21:37 UTC: The disruption runs for about 46 minutes. Anything relying on Claude during this window would have been affected.
- 21:37 UTC: Claude recovers. Anthropic marks the issue resolved.
how a watcher catches this before your users do
NoCrash reads Claude’s public status page once every minute. The moment Anthropic’s own page flips from working to having trouble, NoCrash sends a plain-language message saying what is wrong, in words you can act on, not a raw status code. That means instead of hearing about it from a customer, you hear about it first, within a minute of Anthropic’s own public report, and you can decide what to do: pause a campaign, warn a client, hold off on a send.
It also watches the things you ship. If you have n8n workflows, NoCrash watches those too. If you give it a URL or add a small JS snippet to your app, it watches that side as well. So if your own app goes quiet for a reason that has nothing to do with Claude, that surfaces in the same place. One view, covering the tools you build on and the things you ship.
NoCrash does not find problems before the tool’s own status page does. It reads that page and gets you the information fast, in plain English, in one place.
the authoritative account
For the official record of this outage, go to the Anthropic status page: https://stspg.io/tslzsk6f4hlf