Anthropic’s Claude went down twice on Friday, leaving users unable to reach the chatbot on its website and mobile apps, and disrupting its coding tools and developer services. The first failure struck in the morning, when login errors kept people out of Claude.ai and Claude Code. A second, wider disruption followed in the early afternoon, spreading to the Claude API, its document connectors, and newer services such as Claude Meetings and Managed Agents. Anthropic says the second incident is resolved, with service restored by 15:38 UTC, though some features are still catching up. The company has not said what caused either failure.
The morning failure: logins stopped working
The first signs of trouble appeared in the morning, when Anthropic said it was investigating an issue stopping users from signing in to Claude.ai and Claude Code. Its status page logged the update at 9:58 UTC. Outage tracker Downdetector recorded a sharp rise in complaints around the same time, with about 2,895 reports against a baseline of two.
Claude.ai was listed as a major outage, while the Console and Claude Code showed partial outages and Claude Cowork ran degraded. The API stayed operational through that first incident. Anthropic later deployed a fix for the login problems and marked that incident resolved, before the second disruption began.
The afternoon failure: almost everything degraded
The second incident began at 13:09 UTC, when Anthropic said it was investigating degraded performance across claude.ai, the Claude API, and Claude Code. Within an hour, the list had grown: connectors such as Claude Docs and Claude Sheets, Claude Security, and Claude Managed Agents were all affected, and the mobile apps showed the same problems as the website.
Recovery came in stages. By 14:36 UTC, Anthropic reported improvement on claude.ai and the API, though it warned errors could return while work continued. At 15:33 UTC, errors had returned to normal levels and the company moved the incident to monitoring. Five minutes later, at 15:38 UTC, it marked the issue affecting multiple surfaces as resolved.
Some after-effects lingered. Adding a new connector on claude.ai was temporarily unavailable, scheduled tasks and routines could run more than two hours late, and some runs were skipped. Claude Security scans, reviews, and project setup also remained delayed, while Managed Agents and Meetings were still recovering.

A bad day for workplace tools
The timing made a bad day worse. Microsoft 365 has been down for the same period, with Outlook, Teams, and Copilot disrupted since Thursday, and Anthropic confirmed that its Microsoft 365 connector and Claude for Outlook were caught in that incident. That connector failure began on 12 November, when Anthropic reported elevated errors and traced them to an upstream provider.
Anthropic has not linked the Microsoft 365 incident to Claude’s own two outages, and it has published no cause for either. That leaves the obvious question open: whether the morning login failure and the afternoon degradation shared a root cause, or whether two unrelated faults simply landed on the same day.
Our Read
Two outages in one day is a worse look than one, and the timing could hardly be worse for Anthropic. The company is days away from pricing its public offering, with shares expected to start trading on 19 November, and every reliability wobble gets read as a preview of how it runs infrastructure at scale.
A chatbot going dark is embarrassing; the coding tools and the API going dark at the same time is the part enterprise customers will remember. Claude Code and Managed Agents are sold as infrastructure that work runs on, and a Friday afternoon outage is exactly when that promise gets tested.
The counterweight is speed. The afternoon incident lasted about two and a half hours from first report to resolution, and the morning one was fixed before the second began. Anthropic’s status page stayed updated throughout, with six updates in under three hours.
For a company asking businesses to trust its agents with real work, that candour matters. But two failures in one day will still invite questions until Anthropic publishes a cause, and customers will watch whether the lingering connector and scheduling delays clear cleanly.
Why It Matters
Claude is no longer a side tool for most of its users; it is where writing, coding, and agent work happens. When the website, the apps, the API, and the coding tools fail together, there is no fallback inside the product — users simply stop.
The outage also hit Claude Meetings and Managed Agents, the services Anthropic is pitching to teams, which makes reliability the story rather than a footnote. A service that runs other people’s work cannot afford to be the thing that stops working.
Background
The outage lands in a crowded week for Anthropic, which launched Claude Meetings, opened self-hosted Managed Agents to enterprises, and is in the final days of its IPO roadshow. None of that caused Friday’s failures, but it raises the stakes: the more Claude becomes workplace infrastructure, the more an outage reads as a company event, not a glitch.
Source: Anthropic’s status page and MoneyControl’s report.


0 Comments