Anthropic confirmed a 78-minute outage (17:34-18:52 UTC, Apr 28) hitting Claude.ai, the API, Claude Code login paths, Console, Cowork, and Claude for Government.
Key Takeaways
All major Anthropic surfaces were affected simultaneously: claude.ai, api.anthropic.com, platform.claude.com, Claude Code, Cowork, and Government tier.
Root cause was authentication-layer failures, blocking both browser logins and API token validation for Claude Code.
Incident progressed through four public status stages (Investigating, Identified, Update, Monitoring) over ~25 minutes before resolution was confirmed.
Full recovery was declared at 18:59 UTC; Anthropic stated success rates across all services returned to normal.
Hacker News Comment Review
Enterprise reliability is the dominant thread: multiple commenters report Anthropic’s 90-day uptime has fallen to one 9 (~90%), a threshold most production SLAs treat as unacceptable for a $200k+/month dependency.
The outage is accelerating evaluation of self-hosted models among builders who were already hedging; the authentication scope (hitting Claude Code login paths, not just the web UI) is the specific detail that alarmed operators running automated pipelines.
There is tension between sympathy for the engineering complexity of scaling this fast and frustration that Anthropic support has been slow to respond during incidents, leaving enterprise buyers without a clear escalation path.
Notable Comments
@SimianSci: $200k/month enterprise customer calls Anthropic support “horrendous” and says outage frequency has their exec team “furious.”
@bborud: “When I see people who want to build products on top of these services I can’t help but think that people are mad” – argues instability makes Claude unsuitable as a product dependency today.
