Anthropic’s Claude went offline Monday afternoon. Claude, Claude Code, and related services began experiencing a widespread service interruption around 12:17 PM ET, per Anthropic’s status page — as tracked by TechRadar’s live blog. The failure hit user logins, chat functionality, and code agent execution simultaneously.

Multiple users reported failures across Down Detector. The Economic Times confirmed reports of widespread login and chat failures, with Anthropic’s status page showing “ongoing investigation and recovery efforts” as of the time of publication. No root cause has been disclosed by Anthropic as of this writing.

Which Services Are Down

The disruption covers three surfaces:

  • Claude.ai — the consumer and professional chat interface, with login failures preventing users from accessing sessions
  • Claude Code — Anthropic’s autonomous coding agent, blocked for development teams using it in CI or pair-programming workflows
  • API access — any third-party harness calling the Claude API, including OpenClaw, CrewAI, Autogen, and other frameworks that route to Claude Sonnet or Opus as their model backend

Anthropic has not issued a blanket API blackout notice, but service degradation at the authentication and inference layer typically cascades to dependent tools regardless of whether they hold active sessions.

The Cost of Single-Provider Agent Stacks

Claude is the default or most common model choice across a significant share of the active agent ecosystem. For teams that haven’t built fallback routing to alternative providers, Monday’s outage is a hard stop: the agents don’t degrade gracefully, they simply fail.

The architectural gap is well-understood but widely skipped during initial deployments. Most agent frameworks support multi-provider configuration — routing to GPT-4o, Gemini 2.5, or local models when the primary provider returns errors — but enabling that routing requires circuit-breaker logic most teams don’t write until they’re forced to. An outage of this scope, hitting login, chat, and API in the same window, provides the forcing function.

Teams running time-sensitive pipelines — financial workflows, customer-facing agents, automated coding pipelines — are sitting idle until Anthropic restores service.

Status and What to Do Now

Anthropic’s status page at status.anthropic.com is the authoritative source for recovery updates. As of 18:15 UTC, no estimated time to restoration has been published.

For teams with active deployments: the immediate options are to hot-swap to an alternative model endpoint if the harness supports it, pause scheduled agent runs, or wait for Anthropic’s restoration signal. Teams on OpenClaw can update their model routing config to point at an OpenRouter-available alternative while the outage persists.

This article will be updated when Anthropic confirms a root cause or restoration timeline.