Claude Experiences Outage Affecting API and Apps

Overview

Users of Anthropic’s Claude ecosystem reported significant disruptions on the afternoon of September 29, 2026, with elevated error rates cascading across multiple platforms. The incident affected the primary web interface at claude.ai, the developer-focused Claude API, and specialized tools such as Claude Code and Claude Cowork.

The outage began around 14:00 UTC, prompting immediate investigation by Anthropic’s engineering teams. By 14:36 UTC, initial mitigations were applied, restoring functionality for many existing conversations. However, the company disclosed a secondary issue shortly after that continued to hinder critical user actions, particularly regarding authentication and file management.

This event highlights the increasing reliance on large language model APIs across both consumer and professional workflows, where even brief periods of instability can disrupt productivity for thousands of developers and enterprises.

Timeline of Events

The status of the incident evolved rapidly over the course of approximately twenty minutes before initial stabilizing measures were put in place.

14:21 UTC: Anthropic confirmed it was investigating elevated error rates affecting Claude.ai, including its desktop and mobile applications, as well as Claude Code and Claude Cowork. Users reported failed requests, errors while loading or sending conversations, and unexpected prompts to sign in again.
14:28 UTC: Reports expanded to include degraded performance on platform.claude.com (the Claude Console) and the Claude API at api.anthropic.com. The scope of the outage was confirmed to extend beyond the user-facing chat interface to backend service calls.
14:30 UTC: Engineers acknowledged the broad impact on all major Claude services, including the web platform, console, API, and coding assistants.
14:36 UTC: A mitigation strategy was applied, resulting in a substantial drop in error rates. Existing conversations became mostly functional again. However, a second, distinct issue remained active, preventing many users from completing essential tasks.
15:00 UTC: Updates indicated that while error rates had dropped, failures persisted in sign-in attempts, account actions, and new sessions on Claude Code and Claude Cowork.

"We are working on a fix and will post another update within 30 minutes," Anthropic stated during the height of the disruption, urging users who were already logged in to remain signed in.

Scope and Impact

The outage impacted a wide spectrum of Claude’s services, ranging from casual chat interactions to critical enterprise API integrations.

Web and Mobile Applications

The claude.ai web interface and its accompanying desktop and mobile applications experienced significant degradation. Users encountered difficulties when trying to load past conversations or send new messages. In some cases, the interface appeared to function normally for reading history but failed to process new inputs, creating a confusing user experience.

Developer Tools and API

For the developer community, the implications were more severe. The Claude API, a cornerstone for applications built on top of Anthropic’s models, returned elevated error rates. This likely caused downstream failures in any software dependent on real-time Claude responses.

Additionally, Claude Code and Claude Cowork, tools designed to assist programmers directly within their development environments, were rendered largely unusable. Since these tools rely on persistent connections and frequent API calls, even momentary latency or authentication failures can halt entire coding workflows.

Authentication and Data Integrity

A critical aspect of this incident was the authentication failure. Many users found they could not sign in, particularly those using Single Sign-On (SSO) or Sign in with Apple. This barrier prevented new chat creation and blocked access to existing sessions for a significant portion of the user base.

Furthermore, the inability to upload files posed a specific risk to users who had prepared documents or code snippets for analysis. To prevent data loss or session fragmentation, Anthropic explicitly advised users not to sign out while the team worked on resolving the secondary login issues.

Root Causes and Resolution Efforts

Anthropic has not yet released a detailed technical post-mortem explaining the precise root cause of the dual-issue nature of this outage. However, the pattern suggests two distinct but concurrent failures.

The first issue, related to general request handling and conversation loading, appears to have been mitigated by the 14:36 UTC update. This type of error is often associated with infrastructure scaling limits, database latency, or transient network issues within the model serving stack.

The second issue, which persisted through at least 15:00 UTC, is specifically linked to authentication services. The unavailability of SSO and Apple Sign-In indicates a problem with the identity provider integrations or the token validation systems that manage user sessions.

Current Status

As of the latest update, error rates have decreased substantially, and the core functionality for existing users has been largely restored. However, the pathway to full stability remains incomplete.

Sign-in processes are still intermittent for some users.
New session creation on Claude Code and Cowork continues to face failures.
File uploads remain unavailable for many accounts.

Engineering teams remain focused on implementing a comprehensive fix for the authentication layer. Users experiencing persistent issues are encouraged to check the official status page for real-time updates.

Broader Context

Incidents like this serve as a reminder of the fragility inherent in centralized AI infrastructure. As Claude and similar models become embedded in critical business operations, the cost of downtime extends beyond inconvenience to potential financial and operational losses.

For developers, such outages underscore the importance of implementing retry logic, caching strategies, and fallback mechanisms in applications that depend on LLM APIs. The incident on September 29, 2026, will likely be cited in future discussions about resilience planning for AI-dependent software architectures.

Anthropic is expected to provide a more detailed explanation of the technical failures once the investigation is concluded. Until then, users are advised to exercise patience and avoid unnecessary account actions that might complicate the resolution process.