Crowd Signal Monitor
← Back to tracker

Is The Claude API Down?

This page is for developers trying to tell the difference between a real Claude API incident, a model-specific error spike, and an account- or request-level problem.

The Claude API can fail while the regular Claude web app appears mostly fine, and the opposite can also happen. That is why API problems should be checked as their own surface, not just as part of general Claude downtime.

If you think the Claude API is down, compare these signals:

  • Whether the homepage crowd signal is moving sharply right now.
  • Whether the official status page mentions the API specifically or only claude.ai.
  • Whether one model family is failing while others still work.
  • Whether your failures are global across environments or limited to one account, region, or deployment.

If you are getting isolated 429, 500, or 529 errors without a broader signal, the issue may be request-specific or model-specific rather than a full API outage.

For overloaded capacity specifically, read the Claude 529 error page. For broader sitewide issues, compare this with the live detector on the homepage and the official status lag explainer.

If you are documenting a real incident internally, the most useful details are the affected model, the start time, the error code, and whether claude.ai or Claude Code were impacted at the same time.