Claude not working: the fix for each error message, and how long its last 50 outages lasted

What does your Claude error message mean?

The message itself tells you whose problem it is. A reset time is your plan. A capacity, overloaded or server error is Anthropic's. Find your wording in this table, read 25 October 2026 from Anthropic's error message guide and the Claude Code error reference, then follow the link for the full fix.

Two lines in that table are easy to mix up. A usage limit names a time, because your account has a reset. A capacity message names none, because nothing about your account changed. Claude Code's docs say it plainly: "A 529 is not your usage limit and doesn't count against your quota."

Is Claude down right now, or is it just you?

Open status.claude.com first, but don't treat a green page as proof. The page tracks six parts of Claude: claude.ai, the Console at platform.claude.com, the Claude API, Claude Code, Claude Cowork and Claude for Government. All six read operational at 22:20 UTC on 25 October 2026, a few hours after two incidents that morning.

A green page still leaves one cause open. Anthropic's help center says "Capacity constraints are not outages," and "Capacity issues will not appear on our status page." So a capacity message on a green page is the short kind, which Anthropic says typically resolves "as demand patterns shift throughout the day". Our is Claude down checker shows the status page and user reports side by side.

If the page is green and no capacity message shows, the problem is on your side. Try another browser or the desktop app. If Claude works there, your browser is the cause.

What usually breaks when Claude goes down?

Claude's models break far more often than its logins. On 25 October 2026 we read the last 50 incidents on status.claude.com, which run from 3 August to 25 October 2026, and sorted each one by its title and the parts of Claude it listed. Twenty-eight were errors or slowdowns on Claude models. Two were login failures.

The split decides your first move. When an incident names one or two models, such as "Elevated errors for Claude Opus 5.5" on 24 October, another Claude model usually still answers. When it says "multiple models" or names every service, switching inside Claude won't help, and the wait is longer.

Six of the 50 never touched chat at all. They hit credit purchases, spend limits, Google Play subscriptions and the Console, so claude.ai wasn't among the parts they listed. If you can chat but can't pay or see your usage, that's the kind you're in.

How long does a Claude outage usually last?

A model incident that names specific models lasts about half an hour, and a wider one lasts two to three hours. We timed each of the 50 incidents from its posted start to its resolved notice.

The longest model outage in the window began on 5 August 2026 and ran 429 minutes, a little over seven hours, for "Degraded performance of multiple models". The shortest named one, on 24 October, ended in six minutes.

Why won't Claude load or log you in?

Usually your browser, your VPN or an extension. Anthropic's fix for "There was an error logging you in" is three steps, in its own words:

1. "Ensure you're not using a VPN when accessing Claude."
2. "Disable any browser extensions that you currently have active."
3. "Clear your browser's cache and cookies."

If it still fails, check the status page for a login incident. There were only two in the 50 we counted, both on 24 August 2026, and they ended in 3 and 15 minutes. A text code that never arrives is a different problem, and Claude phone verification covers that one.

Why is Claude Code not working in the terminal?

Run /doctor inside Claude Code, or claude doctor from your shell if it won't start at all. Anthropic's troubleshooting page says /doctor checks your install, settings, extensions and context use, then "proposes fixes it can apply after you confirm."

Then match the error. command not found, a failed install and login loops are in Anthropic's install guide. API Error: 5xx, 529 Overloaded and 429 are in the error reference. A frozen session usually comes back with Ctrl+C. If it doesn't, close the terminal and run claude --resume in the same folder, and the docs say "Restarting doesn't lose your conversation."

Don't retry by hand right away. Claude Code "retries transient failures up to 10 times with exponential backoff before showing you an error", so the error you see has already been retried. For an overload, /model is faster than waiting.

What can you use while Claude is down?

Another Claude model when the incident names one, and another vendor when it doesn't. Copy your unsent message first. In claude.ai, click the model name next to the send button and pick a different one. If the status page says "multiple models", paste a short summary of the thread into GPT or Gemini and keep going. What to use when ChatGPT is down lists the other vendors and their status pages.

Whizi, which publishes this page, keeps Claude next to GPT and Gemini in one chat. When you switch models mid conversation, the thread goes with you (how switching works), and a reply that fails is refunded, not charged. Where it loses: a Claude message sent through Whizi still goes to Anthropic, so an incident across several models takes Whizi's Claude rows down too, and the fix is the same switch to GPT or Gemini. Claude Sonnet and Claude Opus are on Pro at $29.99 a month (Claude in Whizi).