Claude API and Claude.ai Suffer Multiple Outages on April 28

Anthropic logged three separate service disruptions on April 28, 2026, with the most serious one still active as of this writing — knocking out access to both Claude.ai and the Anthropic API simultaneously, a more severe failure pattern than the March 2 outage that spared API users entirely.

The incidents began early in the day when Claude Code Code Review intermittently failed to start new sessions, resolving by 06:00 UTC. A second, shorter episode hit at 13:22 UTC, causing elevated errors before Anthropic declared it resolved at 13:50 UTC — a 28-minute window that caught developers mid-session. Then at 17:41 UTC, a third and more significant incident emerged: users found they could not reach Claude.ai at all. At 17:51 UTC, Anthropic expanded its assessment to include the Anthropic API itself, which was also showing elevated errors and blocking Claude Code login paths.

As of the time of writing, the third incident remains open on Anthropic's status page with no resolution timeline disclosed.

More Severe Than March

The March 2, 2026 outage that hit thousands of developers during the US East Coast workday had one key saving grace: the Claude API remained fully operational throughout. Enterprise developers and teams using direct API integrations were largely insulated from that disruption.

April 28 is a different story. Anthropic's status page states the current incident involves "elevated errors on the Anthropic API, as well as issues accessing Claude.ai, including log-in paths for Claude Code." That means developers who avoided the consumer-layer disruptions in March are now directly in the blast radius. API-dependent workflows, CI/CD pipelines that call Claude, and automated agents built on the Anthropic API are all hitting elevated error rates.

Downdetector and third-party monitoring service IsDown both registered the issue immediately. IsDown tracked 1,744 user reports in the 24-hour period as of 2:49 PM EDT on April 28, with active reports still coming in. Other trackers placed peak report counts above 5,000 across the day's combined incidents.

A Pattern of Instability

Today's triple-incident day is not an anomaly in isolation — it lands in an already-troubled reliability window. According to IsDown's historical tracking, Claude has logged 137 incidents in the past 90 days, split between 43 major outages and 94 minor incidents, with a median resolution time of 1 hour and 8 minutes per incident.

The Register reported that developer complaints about Claude Code quality have been rising since January 2026, not just availability. When pointed at Claude Code's own GitHub issue tracker, Claude itself assessed: "Yes, quality complaints have escalated sharply — and the data tells a pretty clear story." Anthropic traced part of the quality degradation to three separate code changes that affected Claude Code, the Agent SDK, and Claude Cowork, which it says were fixed as of April 20. The current outages are separate from that quality issue.

The reliability pressure appears tied to rapid scale. Anthropic has shipped significant new products in recent weeks — Claude Opus 4.7, Claude Design (an Anthropic Labs product), and Claude Managed Agents in public beta — all while its infrastructure absorbs mounting demand from an expanding enterprise customer base. Whether today's incidents stem from that growth pressure or a distinct infrastructure failure is not yet clear.

What's Unconfirmed

Anthropic has not provided a root cause for any of the three April 28 incidents. It is unclear whether they share a common origin or represent separate infrastructure failures. No incident credits or post-mortem timeline have been announced. Whether the current active outage shares a failure mode with the resolved 13:22 UTC episode is also unconfirmed.

For developers running API-dependent workloads or Claude Code in CI pipelines, the current incident warrants checking status.claude.com directly. Teams that were reassured by the March 2 incident's API-only-spared outcome should note that today's scope is broader — both the API and consumer surfaces are affected simultaneously for the first time in recent memory.