Is Claude Down? How to Check in 30 Seconds (and Keep Your App Up)

Is Claude down? Check status.claude.com (external) first. It shows separate status for claude.ai, the Claude Console, the Claude API, Claude Code and more. Then read your error: 529, 500 or 504 points to Anthropic, while 401 or 400 usually means something on your side.

By Gary, Greenlight My App · Updated Oct 30, 2026 · How we checked this: Tested on Oct 30, 2026, 7:00–7:11 AM ET. We read status.claude.com, sent one API request with a fake key, built a small demo app with a leaky and a safe error page, and scanned it with Greenlight My App before and after the fix. We also read Anthropic's API errors documentation and Downdetector. We haven't tested Claude Code for this article. Drafted with AI tools and reviewed by a person before publishing.

This page doesn't show live status. It shows you how to find out in 30 seconds, and how to make sure your own app doesn't fall over, or leak details, the next time Claude has a bad hour.

Is Claude down right now? How to check in 30 seconds

Three steps, in this order:

1. Open status.claude.com (external). Look at the top banner and at the component you care about. When we checked at 7:10 AM ET on Oct 30, 2026, the page said "All Systems Operational" and listed claude.ai, Claude Console (platform.claude.com), Claude API (api.anthropic.com), Claude Code, Claude Cowork and Claude for Government, each with its own 90-day uptime.
2. Read your error code. See the table below. It's the fastest way to tell an outage from a setup problem.
3. Cross-check user reports. Downdetector (external) shows a chart of user reports. On Oct 30, 2026 it said user reports showed no current problems.

The 90-day uptime figures shown that morning: claude.ai 99.71%, Claude Console 99.84%, Claude API 99.76%, Claude Code 99.71%, Claude Cowork 99.7% and Claude for Government 100%.

Which Claude component is down?

Outages don't always hit everything. Claude.ai, the API and Claude Code can fail separately.

Real example: on Aug 16, 2026, BleepingComputer reported a major outage starting around 21:58 UTC (5:58 PM ET) that hit Claude.ai, Claude Code and Cowork with sign-in problems, while the Console and API were listed as operational. Service was back by about 22:40 UTC (6:40 PM ET) (BleepingComputer (external)). If you just watched the top banner, you'd have missed that the API was fine.

What do Claude API error codes mean?

Anthropic's errors documentation (external) lists these codes. Here's who usually needs to act.

Two details worth knowing from the docs:

• A 429 caused by a spend cap doesn't include a retry-after header. Retrying won't help; raise the cap or wait.
• Every response includes a request-id header. Log it. Anthropic support will ask for it.

Worked example: your chatbot starts failing. Your logs show overloaded_error with status 529 on most calls, and status.claude.com has an open incident for the API. That's an outage. Show a friendly message and retry later. If the logs instead show authentication_error with status 401 and the status page is all green, someone probably rotated or revoked the key. See our Claude API key security guide.

Is Claude down, or is it my API key?

If the status page is green and you're getting 401s, check these:

• Is the key set in the environment the app actually runs in? A key in your local .env file doesn't reach production by itself. Our what is a .env file guide covers this.
• Was the key revoked? If it was ever committed to git or shipped in frontend code, it may have been disabled. Read is my AI API key stolen?.
• Is the key the right shape? Anthropic keys start with sk-ant-. A placeholder like sk-ant-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX left in config will fail every time.

Tested: on Oct 30, 2026 at 7:10 AM ET we sent a real request to the Claude API with a fake key (sk-ant- followed by 40 X's). The status page was all green. The API answered HTTP 401, authentication_error, "invalid x-api-key". That's the pattern to recognise: a 401 while the status page is green means it's your key, not an outage.

If you're getting 429s, check your rate limits and spend settings in the Console, and see Claude API pricing for how usage adds up.

How do I keep my app up when Claude is down?

You can't fix Anthropic's outage, but you can make yours shorter and quieter.

Retry with exponential backoff

Short blips often clear in seconds. Anthropic's docs say the official SDKs retry some transient errors twice by default with exponential backoff, honour the retry-after header, and let you change this with max_retries. If you call the API directly, add the same logic: wait, then double the wait each time, with a cap.

// Runs on the server. The key comes from the environment, never the browser.
import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  apiKey: process.env.ANTHROPIC_API_KEY,
  maxRetries: 3,
});

Add a timeout and a fallback

Set a timeout so users don't wait forever. For features that matter, fall back to a cached answer, a simpler path or another model provider. Decide this before the outage, not during it.

Show a safe error page

When the API fails, users should see something like "Our assistant is having trouble right now. Try again in a few minutes." They shouldn't see the raw error.

Raw errors can reveal which provider and model you use, internal file paths and sometimes headers. That's free reconnaissance for attackers.

Tested: we built two versions of a small "Ask Claude" demo and hit the same real 401 in both. The leaky version called the API straight from the browser. Its error page showed a stack trace, the raw API error and the request headers, including the (fake) API key. The safe version sent the browser's request to its own server route. Users saw "Our AI assistant is unavailable right now. Please try again in a few minutes," with a reference ID, and the full 401 detail stayed in the server log.

We checked the error pages by eye from these screenshots. Greenlight My App's scanner doesn't look for stack traces on error pages. What it does check is whether a key is in your public JavaScript:

• Before: the scan returned FAIL, "Anthropic API key is visible in your public code".
• After moving the call to a server route: PASS, "No secret keys found in your public code".

Cropped to the relevant result.

Get alerts instead of finding out from users

status.claude.com offers notification options, including webhooks, so you can hear about incidents automatically. Pair that with an alert on your own error rate.

How often does Claude go down?

Status pages show recent incidents and uptime history, which is the honest answer. On Oct 30, 2026, status.claude.com showed 90-day uptime per component (listed above), no incidents on Oct 29 or 30, and a resolved Google Calendar connector incident on Oct 28. Most incidents are partial: one model, one feature or one product, not everything at once. That's why checking the specific component matters.

For founders: what else breaks when an AI API fails?

An outage is a stress test. Things that commonly go wrong in AI apps:

• Raw provider errors shown to users.
• API keys in frontend code, so a "fix" during the outage gets copied to the browser.
• No server-side rate limiting, so retries pile up and run up a bill when service returns.

If your app is built with Claude Code, our Claude Code permissions and sandbox post covers related setup. Run a free Greenlight scan to check your live app's public code for exposed keys.

FAQ

Is Claude down right now? This page doesn't show live status. Check status.claude.com for the official view, and Downdetector for user reports. If both look normal and you still get errors, read your error code; 401 and 400 usually point to your key or request.

What does Claude error 529 mean? 529 is overloaded_error. Anthropic's docs say it means the API is temporarily overloaded because of high traffic across all users. It's not caused by your key. Retry with backoff, and show users a friendly message if it continues.

Is the Claude API down when claude.ai is down? Not always. They're separate components on the status page. In the Aug 16, 2026 outage reported by BleepingComputer, Claude.ai, Claude Code and Cowork had problems while the Console and API were listed as operational.

Did you test Claude Code for this article? No. We haven't tested Claude Code for this article. The Claude Code row in the status table comes from status.claude.com, not from our own testing.

Want to know what your users see when an API fails? Run a free Greenlight scan (external) to check for exposed keys in your public code and missing security headers.