Anthropic's API and claude.ai saw elevated errors for 28 minutes on 8 November, starting 08:28 UTC. Anthropic traced it to Claude Fable 5.1 and resolved it before 09:00 UTC. The fault sat on Anthropic's side.
Started
08:28 UTC
Duration
Lasted 28m
Source
IsDown
Next time
Anthropic is degraded. Is your product working?
An incident on their side does not always mean an outage on yours, and the only way to know is to already be watching your own endpoints. Set that up in under a minute and you will be notified the next time your own site is affected. And there will be a next time.
50 monitors for your own websites. No card.
Affected components: api.anthropic.com, claude.ai
What happened?
Anthropic logged one incident on 8 November 2026, starting at 08:28 UTC and lasting 28 minutes. The title Anthropic gave it was "Degraded performance for Claude Fable 5.1", and the components it listed as affected were api.anthropic.com and claude.ai. Anthropic's first update said it was investigating elevated errors on requests to Claude Fable 5.1. Eleven minutes later it posted that a fix had been deployed and it was monitoring for recovery. The resolved update, posted at 08:54 UTC, widened the description: it said the issue affected claude.ai, the Claude API, Claude Code, and Claude Cowork, and gave the actual impact window as 08:18 to 08:32 UTC, narrower than the span between the first and last update. No root cause was given in any of the three updates, so why the errors happened is not on the record. There were no user reports filed during this window. Over the past 90 days Anthropic has logged 46 incidents with a median duration of 55 minutes, so this one resolved faster than typical. Across all time Anthropic has filed 849 incidents, averaging 29.3 per month. Anthropic has not published a post-mortem for this incident.
Learning
Anthropic's own status updates show a 16-minute gap between when it says impact actually started, 08:18 UTC, and when it opened the investigating update, 08:28 UTC. That gap matters if you were relying on your own endpoint checks to catch this: a check hitting your product's surface would only have seen errors once Anthropic's underlying models were already degraded, and by the time Anthropic confirmed the problem, it was already being fixed. This kind of incident sits entirely on the provider's side, in model-serving infrastructure your monitoring never touches directly, so a clean result from your own health checks during this window would not have been wrong, it would just have been answering a different question. The pattern here, elevated errors followed by a deployed fix within minutes, is typical of Anthropic's 90-day history, where the median incident runs 55 minutes. Your monitoring covers your half, UptimeRobot now watches the provider's half too, so next time you know which side broke without guessing.