Claude Outage History

The Filesystem connector was missing from Claude Desktop, preventing users from accessing file system functionality. The issue affected claude.ai for 46.5 hours until it was identified and fixed. Users needed to re-add the Filesystem connector to their organization allowlist to reactivate the feature after the resolution.

Network degradation at an upstream peering point caused TCP handshake failures and connection timeouts for some API requests to api.anthropic.com, affecting Claude.ai service. API requests that successfully established connections continued to work normally. The issue was resolved after 48 minutes by implementing a fix and working with upstream providers.

Claude experienced elevated errors with Haiku 4.5 that affected both claude.ai and api.anthropic.com for 1 hour, with users also reporting issues with Opus 4.6 and Sonnet 4.6. The service team implemented a fix and monitored the results until errors returned to baseline levels.

Claude Haiku 4.5 experienced elevated errors for 5 minutes between 14:45-14:50 UTC on March 4, 2026. The issue affected both claude.ai and api.anthropic.com services. The incident was resolved and service was restored to normal operation.

Claude Opus 4.6 experienced elevated errors for 1.7 hours, affecting both claude.ai and api.anthropic.com. Users reported that every prompt was returning API error 400, preventing normal service usage. The incident was resolved after investigation and monitoring.

Claude Sonnet 4.6 experienced elevated errors for 56 minutes, affecting both claude.ai and api.anthropic.com. Users reported that the AI was not responding to any requests during the incident. The issue has been resolved.

Claude Haiku 4.5 experienced elevated errors for 48 minutes, affecting both the claude.ai website and api.anthropic.com API service. The incident was resolved after investigation and remediation efforts.

Claude Opus 4.6 experienced elevated errors for 7.5 hours, affecting claude.ai, console.anthropic.com, and api.anthropic.com. The issue was intermittent, recurring after initial fixes were attempted throughout the day. The incident was ultimately resolved after multiple fix implementations and monitoring cycles.

Claude experienced elevated errors across claude.ai, cowork, platform, and claude code services for 7 hours, affecting both the main claude.ai interface and the console.anthropic.com platform. A fix was implemented and monitored before the incident was fully resolved.

Claude experienced elevated errors specifically affecting the Haiku 4.5 model, impacting claude.ai, console.anthropic.com, and api.anthropic.com for 3.3 hours. Users reported document upload failures and very slow response times with no output, particularly affecting Opus 4.6. The issue required multiple fix attempts before being fully resolved.