Claude went down for two hours on 13 November with sign-in errors

Anthropic's status page recorded a two-hour incident on 13 November. The first entry says Anthropic was investigating a problem with users signing in to claude.ai and Claude Code. The page lists four affected services at the bottom. They are claude.ai, Claude Console at platform.claude.com, Claude Code and Claude Cowork.

The Hacker News thread drew 488 points. That put it near the top of the front page that day.

What the status page says

The first entry is dated 09:58 UTC. It says Anthropic was looking at a problem with users signing in to Claude.ai and Claude Code. The 11:05 UTC entry says a fix was deployed, and Anthropic was watching for recovery. The 11:58 UTC entry says the problem affecting claude.ai and Claude Cowork was resolved.

Three updates, no further detail. Anthropic has not written down a cause. There is no review of the incident yet. The two-hour window is what Anthropic confirms. Anything about what broke is guesswork.

What was and was not said

The earlier outage on 29 September carried a note about messages. Some messages sent in the window may not have been saved. This page carries no such note. The only service actions listed are the fix at 11:05 UTC and the resolution at 11:58 UTC.

Anthropic listed four services. claude.ai is the chat site. Claude Console at platform.claude.com is where teams manage their API keys and billing. Claude Code is the coding agent that signs in to the same account. Claude Cowork is the meeting and workspace tool. All four share the sign-in path. That is what the first entry names as the problem.

What a team should check

For a team that runs scheduled Claude Code jobs, start with the 09:58 to 11:58 UTC window on 13 November. A scheduled run that failed to sign in is the easiest case. The log shows the sign-in error. The next scheduled run picks up.

A run that signed in before 09:58 UTC is the harder case. It may have tried to call the API inside the window. The call may have returned a retryable error. The job may have stopped halfway.

The second thing to check is any script that signs in through Claude Console. A key rotation or a billing change in flight during the window may need to be done again.

Shared sign-in keeps failing in the same shape. One sign-in path gates every product. When the path breaks, every product that depends on it stops. A team that runs production work against one model vendor is one incident away from a quiet morning. A tested fallback path to a second model pays for itself the first time it is used.

Source

Primary source: Elevated errors on claude.ai, Claude Code, and platform.claude.com on the Anthropic status page.