Guides

How to tell if Claude is down or if it's just your account

Outage, usage limit or sign-in fault? How to tell which one has stopped Claude, with Anthropic's 29 September 2026 outage as the worked example.

Multi Chats Team
November 7, 2026 · 6 min read

When Claude stops answering, what exactly does your screen say?

Those few words usually settle it. An outage, a used-up allowance and a sign-in fault look the same from your chair, and each one has its own fix. You can normally tell them apart in about a minute.

Three things that look like "down"

Here is how each one shows itself.

  • A real outage. Anthropic's systems are failing for many people at once. It appears on the official status page, often after a delay.

  • Your usage limit. The service is fine, but you have used your allowance for now. Claude says so. According to Anthropic's help centre, the blocking message reads "5-hour limit reached - resets [time]."

  • A sign-in problem. Claude is up, but you cannot get into your account. The generic version is "There was an error logging you in".

A fourth case is milder. When demand is high, Claude can reply "Due to unexpected capacity constraints, Claude is unable to respond to your message. Please try again soon." Anthropic treats this as normal load management, and says these episodes "will not appear on our status page". Its advice is to try again in a few minutes.

MultiChats puts the leading AI models in one app. Try it free

Check it in about a minute

Go in order. Stop when one gives you the answer.

  1. Read the message. A reset time means your limit. The capacity line means wait a few minutes. In both cases nothing is broken.

  2. Open status.claude.com. It gives claude.ai, the API and Anthropic's other products a line each. Look at the line for what you use.

  3. Start a new chat and send one word. If it answers, the old conversation is the problem. Claude itself is fine.

  4. Try your phone on mobile data. Or the browser, if you were in the app. A different device on a different network rules out most local causes in one go.

Still failing on a second device and a second network? Then the fault is very likely Anthropic's, even if its status page has not caught up.

29 September 2026, hour by hour

Anthropic's own record of that afternoon shows how this plays out.

Errors began at 14:00 UTC (16:00 in Brussels). They hit claude.ai, the Claude apps, Anthropic's other products and its API. The first post on the incident page went up at 14:21. For those first 21 minutes, people were stuck and nothing had been posted.

Anthropic described two problems. The errors were mitigated at 14:36 UTC, and existing conversations mostly worked again. A second issue kept many people from signing in. Single sign-on and Sign in with Apple were unavailable. Starting new chats, voice conversations, purchases and file uploads failed too.

At 15:00 UTC came the most useful line of the day: "If you're signed in, please don't sign out." With sign-in broken, anyone who signed out to "reset" things risked being unable to get back in.

At 15:11 UTC, Anthropic said most services had recovered at 14:59. It added one warning: "Some messages sent between 14:00 and 14:59 UTC may not have been saved." The incident was marked resolved at 16:27.

What that hour teaches

Stay signed in during an incident. If sign-in is part of the problem, signing out can leave you on the wrong side of the door.

Keep your own copy of anything long you send while things are shaky. Anthropic's warning covered a whole hour of messages.

Why the status page and your screen can disagree

There are three ordinary reasons.

  • It lags. Someone has to confirm an incident and write it up. A calm page in the first few minutes proves little.

  • Busy periods are left off. Capacity messages are not listed, by Anthropic's own account.

  • Sometimes only some models break. On 22 September 2026, from 00:50 to 02:10 UTC, the errors Anthropic reported were on three named models: Claude Mythos 5.1, Claude Fable 5.1 and Claude Opus 5 (incident page). On one of those, Claude looked down. On another model, you may have noticed nothing.

Our Claude status page reads Anthropic's official feed every three minutes and lists recent incidents with dates. Our main AI status page also watches our own traffic. A model appears in its Model Availability list only when the requests we sent it in the past 30 minutes failed. That list can show a problem while Anthropic's page is still green.

Your next step, case by case

If it is an outage, wait, and stay signed in. Going by the status page's own timestamps, every September 2026 incident that touched claude.ai was marked resolved within three hours of its first post, and several within one.

If it is your limit, the message tells you when it resets. Anthropic's usage guide says that on a paid plan such as Pro or Max, Settings, then Usage shows progress bars for your five-hour session and weekly limit. If you keep hitting limits, our usage limits comparison sets out how each big assistant counts them.

If it is sign-in, Anthropic's error guide suggests turning off any VPN, disabling browser extensions and clearing your cache and cookies. If you log in by email and the link never arrives, its login page says to check spam for a message from @mail.anthropic.com. Then look at the status page, because on 29 September sign-in was part of the outage.

Using Claude inside MultiChats?

Then an Anthropic outage reaches you too. We connect to Claude through Anthropic's API, and on 29 September Anthropic reported that some API requests were failing. When Claude is down at Anthropic, it is down in our app, and we cannot route around that.

Two things work differently.

  • Your account is ours. You sign in to MultiChats with our own login, so a Claude sign-in fault does not lock you out of your chats here.

  • The chat can carry on. You can switch the same conversation to a model from another company and keep going. Pick one in the model picker before your next message, or open the regenerate menu under a reply, which lists other models to retry with.

Gemini 3.8 Flash and GPT-6.1 Sol both sit on the same paid plans as Claude. GPT-6.1 Sol has no web search of its own in our app, so take questions about today's news to a model that does, such as Gemini. If a very long chat will not fit in the model you switch to, the app tells you and offers to switch back. During an outage that will not help, so start a new chat with a short summary instead.

Once Claude is back

Switch the chat back to Claude whenever you like. What you wrote with the other model stays in the conversation.

If you were in Claude's own app during an outage, scroll up and check your last messages made it. On 29 September, Anthropic could not promise they all had.

For the same check on the other big assistant, see what to use when ChatGPT is down. Incident times above come from Anthropic's status page, checked on 2 October 2026.