Claude.ai, Claude Code, and Claude Cowork were disrupted for roughly 36 minutes on August 16, starting as an authentication issue and widening into degraded performance across Anthropic's core surfaces. Anthropic's status page shows the incident as fully resolved; the company has not disclosed a root cause.
Anthropic first flagged authentication problems at 21:58 UTC on August 16. At 22:02 UTC it broadened the report to degraded performance on claude.ai, platform.claude.com, the Claude API, Claude Code, and Claude Cowork. A fix was deployed by 22:22 UTC; Anthropic marked all affected services restored at 22:34 UTC.
What's new
- Aug 16 timeline (Anthropic status page): auth investigation 21:58 UTC → degraded performance 22:02 UTC → monitoring after fix 22:22 UTC → resolved 22:34 UTC.
- Root cause: Not disclosed.
- 90-day uptime (Anthropic status, current): claude.ai 99.35%, Claude API 99.46%, Claude Code 99.38%, Claude Cowork 99.47% — below the 99.9% threshold common in enterprise SLAs.
- August density: Anthropic's own incident log shows multiple degradations in the same week — including a multi-hour Claude Fable 5 episode on Aug 14–15 (impact logged from 20:00 UTC Aug 14 to 00:11 UTC Aug 15) and a roughly four-hour multi-model disruption on Aug 12 (13:50–18:07 UTC).
- Third-party counts: Monitoring aggregator StatusGator logs 173 Claude outage events since January 2026, as of this writing — a higher-frequency count that includes partial degradations its sensors flag independently of Anthropic's major-outage labels.
Why it matters
A single 36-minute disruption is not, on its own, catastrophic. What makes August worth tracking is frequency: Anthropic's status history shows several multi-hour degradations across Claude.ai, the API, Code, and model-specific error spikes in the same week — not one isolated auth bug. Enterprise contracts built around 99.9% uptime typically budget about 43 minutes of downtime per quarter; Anthropic's published 90-day figures on claude.ai already imply several hours of recorded impact across the period.
As more developers route production agentic workloads through Claude Code and the API, each degradation shifts from a login inconvenience to measurable workflow risk — especially when auth, API, and Code fail together rather than on a single surface.
Our take
Anthropic has been transparent on status-page timelines, but transparency does not equal stability. The open question is whether compute scaling can keep pace with demand before uptime — not benchmark scores — becomes the metric enterprise buyers weigh most heavily.
