AI APIs
Is Anthropic down?
Live Anthropic status, measured independently. We probe 5 Anthropic endpoints from monitoring regions on every continent and record every outage we detect.
Anthropic is up
Our probes are reaching Anthropic normally from every region we check from.
Checked every minute from 6 regions on every continent.
- 24h uptimeUptime, 24 hours
- 7d uptimeUptime, 7 days
- 30d uptimeUptime, 30 days
- API100% uptime, 30 days6 regions, 97 to 247 msCheckingp50184 msMedian responsep95246 msSlowest 5% of checks500 ms250024 h ago12 h agonow
- Console99.94% uptime, 30 days6 regions, 266 to 640 msCheckingp50475 msMedian responsep95942 msSlowest 5% of checks1000 ms500024 h ago12 h agonow
- Documentation99.96% uptime, 30 days6 regions, 523 to 1283 msCheckingp50980 msMedian responsep951754 msSlowest 5% of checks2000 ms1000024 h ago12 h agonow
- Website100% uptime, 30 days6 regions, 120 to 365 msCheckingp50245 msMedian responsep95365 msSlowest 5% of checks500 ms250024 h ago12 h agonow
- Official status feed99.86% uptime, 30 days5 regions, 213 to 487 msCheckingp50333 msMedian responsep95519 msSlowest 5% of checks500 ms250024 h ago12 h agonow
Daily strip:99.5% or better99–99.5%95–99%below 95%no data yet
Get notified when Anthropic goes down
Follow these 5 checks from your own workspace and get alerted through Slack, email, webhooks, or on-call the moment Anthropic goes down. They don't count against your monitor quota. Free plan, no card required.
About Anthropic outages
The Anthropic API serves Claude models to products that use them for chat, coding, analysis, and agents. Like every model provider, incidents usually appear as elevated error rates or latency on specific models rather than a full outage, and they can be region-specific.
What we check
Every row is a real monitor in Uptimy, running from each of our regions. A failure in some regions marks the row degraded; failures from every region mark it down and open an outage.
| Check | Target | Expects | Interval |
|---|---|---|---|
API The Anthropic API is not answering with its normal authentication response. Messages requests fail. | httpapi.anthropic.com/v1/models | HTTP 401 | 1 min |
Console The Anthropic Console is unreachable. API keys and usage dashboards are affected. | httpconsole.anthropic.com | HTTP 200 | 2 min |
Documentation Anthropic's developer documentation is unreachable. | httpdocs.claude.com | HTTP 200 | 5 min |
Website anthropic.com is unreachable. | httpwww.anthropic.com | HTTP 200 | 2 min |
Official status feed Anthropic's status page is unreachable. | httpstatus.claude.com/api/v2/status.json | HTTP 200 | 5 min |
For component-level detail and incident write-ups from the vendor, see Anthropic's official status page.
Embed a live Anthropic badge
A status badge for your README, wiki, or runbook that updates every minute from these probes and links back to this page. Pick the whole service or a single check.
Preview
Markdown
[](https://www.upti.my/status/anthropic?utm_source=badge)HTML
<a href="https://www.upti.my/status/anthropic?utm_source=badge"><img src="https://www.upti.my/status/anthropic/badge.svg" alt="Anthropic status"></a>Recent Anthropic outages we detected
An outage is recorded when a probe fails from every monitoring region and closed when it recovers. Partial and single-region problems show as degraded above but are not listed here.
Anthropic is up. Is your Anthropic integration?
The incidents that actually hurt are usually not global outages. They are the parts of your own system that depend on Anthropic and fail quietly. Here is what teams that rely on Anthropic monitor on their side.
Probe a real request from your backend
Reachability is not availability for model APIs. Run a small request through your own service and alert on latency and error rate.
HTTP monitoring with latency alertsAutomate the fallback
A workflow can detect provider errors, switch to a fallback, and post to your status page without a human in the loop.
Workflow monitoring and automationAnthropic status: frequently asked questions
An overloaded error means capacity pressure on a model, which the official page reports as degraded performance. It is not the same as the API being unreachable, which is what our probe measures. Retry with backoff and consider a fallback model.
Look at the per-region results above. If our probes fail from every region, Anthropic has a real outage and it will appear in the outage list. If they succeed everywhere, the problem is most likely on your side: your network, DNS resolver, a browser extension, a corporate proxy, or your own integration. If only some regions fail, Anthropic may have a regional incident that does not affect you.
Every probe runs at least every five minutes from each monitoring region, and most run every minute. This page refreshes its data every minute. Uptime figures are calculated from all checks across all regions.
Official status is updated by people after an incident is confirmed and often lags the first failures by many minutes; our probes react within a minute or two. In the other direction, we only see public endpoints, so an internal incident may not show on our probes. Use both.
Yes. Use "Notify me about Anthropic" above: your Uptimy workspace follows these exact checks, free plan included, and every outage opens an incident there and routes alerts to Slack, email, webhooks, or on-call. Nothing is duplicated and it doesn't use your monitor quota. Better still, also monitor your own dependency on Anthropic: the API route, webhook receiver, or scheduled job that calls it, so you learn about the failure that actually affects your users.
Related
Other ai apis
Uptimy is not affiliated with Anthropic. Anthropic and related marks are the property of their respective owners and are used here only to identify the service being measured. Results reflect the reachability of public endpoints from our monitoring regions and may differ from the vendor's own reporting.