deniz.in

Markets

Weather

Loading weather

· via dev.to (home feed)

Claude hit by two outages on 13 November as Anthropic hands paid users a one-time limit reset

Claude suffered two separate outages on 13 November 2026 — one breaking sign-in, another degrading the API — while Anthropic gave Pro, Max, Ultra and Team accounts a one-time free limit reset that expires on 10 December.

Claude hit by two outages on 13 November as Anthropic hands paid users a one-time limit reset

Two outages on 13 November

Anthropic's Claude went down twice on 13 November 2026, according to incident data from Anthropic's status page compiled in a timeline post on dev.to. The first outage, from 09:58 to 11:58 UTC, broke authentication for claude.ai and Claude Code, with many users staring at the error "403 Attention Required! | Cloudflare". The second, from 13:09 to 15:38 UTC, caused errors across claude.ai, the Claude API, Claude Code, Claude Cowork and connectors. Both are resolved, but Anthropic has not published a cause for either.

What the sign-in outage broke, and what stayed up

During the morning incident, claude.ai was in major outage, Claude Code and the Console in partial outage, and Claude Cowork degraded. The Claude API, Claude for Government and Claude on Cloud Platforms (AWS, Google Cloud, Microsoft Foundry) stayed operational — so an API key or a cloud-provider deployment kept you working. Downdetector logged roughly 3,800 reports around 10:15 UTC, according to The Sun as cited in the timeline. A fix went out at 11:05 UTC and the incident closed at 11:58 UTC.

The dev.to post explains what the 403 actually meant: Claude Code's request came back with Cloudflare's browser-challenge page instead of a token, and a terminal tool cannot complete a browser challenge, so it surfaced as an authentication failure. It also hit mid-session, not just at sign-in, knocking users out of running tasks.

The afternoon incident hit the API too

The second incident started as degraded performance on claude.ai, the API and Claude Code, then widened to connectors (Claude Docs, Claude Sheets), Claude Security and Claude Managed Agents. Claude.ai and the API recovered by 14:36 UTC; scheduled tasks and routines ran more than two hours late or were skipped before the incident closed at 15:38 UTC. Because the API itself was affected, an API key offered no escape route this time — the inverse of the morning.

The Microsoft 365 connection was mostly a red herring

A separate incident, from 16:03 UTC on 12 November to 16:40 UTC on 13 November, hit only the Microsoft 365 connector and Claude for Outlook, and Anthropic attributed that one to a Microsoft 365 service incident. Some headlines tied the Claude outages to Microsoft; Anthropic's status page does not, and the dev.to author notes a developer on Hacker News with no Microsoft dependency was logged out of Claude Code the same way. The one real overlap: Claude login codes mailed to Microsoft 365 mailboxes may never have arrived.

Anthropic's one-time free limit reset

Separately, on 10 November 2026, Anthropic gave every Pro, Max, Ultra and Team account one free full limit reset, according to a second dev.to post. It must be claimed by 10 December 2026 at 2:00 PM PT (22:00 UTC) via Settings > Usage in the Claude apps. A full reset refills every limit across all models immediately, but it does not shift the weekly reset day, does not raise limits, does not refund billed overage, and cannot be undone. Limits are shared account-wide, so restored usage also counts in Claude Code. Free plans get no resets, and several Enterprise configurations are excluded.

The hidden /limit-reset command

Claude Code ships an undocumented /limit-reset command accepting full, weekly, 5-hour or fable as arguments, but it only works when Anthropic enables it for your account; otherwise it answers "Unknown command". The Settings > Usage button does the same job.

Anthropic's stated motive, quoted from a ClaudeDevs announcement, was that it "recently fixed a few small issues that used more limits than intended", and that Fable-related resets had done little for people already at their weekly caps. GitHub issues in the same period described a Team account hitting its weekly limit about 30.5 hours after reset, faster drain since mid-October, and headless runs consuming limits about 1.8 times faster than interactive ones — but those are user reports, and Anthropic has not confirmed which problems it fixed.

Timing decides the value: spent early in the weekly cycle a reset buys nearly a second week; spent hours before the scheduled reset it buys almost nothing. Unused resets expire, and cancelling a plan before using one forfeits it.

Why it matters

For anyone whose workflow runs through Claude Code or the API, the 13 November incidents show that two outages on the same day can need opposite workarounds: an API key bridges a sign-in failure but not an API failure, so the status page — not habit — should decide your response. The free reset is effectively compensation for metering bugs, and it comes with a hard 10 December deadline that paid users who hit limits regularly should not let lapse. Both stories also highlight a transparency gap: days later, neither outage had a published root cause.

  • #claude
  • #anthropic
  • #outage
  • #claude-code
  • #api

Related posts