Anthropic’s AI service Claude experienced another large-scale anomaly on April 15th, Eastern Time. The impact was not only on the chat interface Claude.ai, but also extended to Claude API, Claude Code, and even Claude Cowork was reported to be unable to function normally. According to Anthropic’s official status page and multiple foreign media reports, the incident took from investigation, confirmation, repair to final recovery, ranging from tens of minutes to more than an hour, which had a significant impact on users who relied on Claude for writing, research, and program development.
Judging from the nature of the incident, this time it was not simply a stuck entry page, but a systemic failure across multiple core services. Anthropic clearly wrote on the official incident page that they observed an increase in the error rate of Claude.ai, API and Claude Code. They later updated to explain that Claude.ai and platform services were completely interrupted for a time, and Claude Code was also affected because the login process relied on Claude.ai. As of subsequent updates, Anthropic stated that the API has been restored first, but the Claude.ai error is still being processed. Although logged-in Claude Code users can continue to work, there are still problems with the new login process.
Official confirmation: Claude.ai, Claude Code, and API all have problems
An Anthropic spokesperson pointed out that Claude.ai and Cowork were unavailable to most users that afternoon, and some users were unable to log in to Claude Code. Anthropic stated that the team is actively repairing the problem and thanked users for their patience while the service was restored.
Anthropic’s own incident page provides a more complete timeline. According to records, the official announcement was made for the first time at 14:53 UTC on April 15 that it was investigating the increase in errors in Claude.ai, Platform, Claude API and Claude Code. At 15:03 UTC, it stated that the problem had been found and began to import fixes. An update at 15:20 UTC stated that the API had been fully restored, but problems related to Claude.ai and login were still being eliminated. At 15:40 UTC, it was clearly stated that Claude.ai and Platform were once down. Status, Claude Code’s login mechanism through Claude.ai also failed to work; it was finally announced at 17:42 UTC that the incident had been resolved.
External monitoring data shows that the number of error reports has soared rapidly.
In addition to the official status page, external user reports also reflect the scale of this anomaly. Mashable noted that user error reports on DownDetector spiked rapidly around 10:30 a.m. ET, with 38% of reports related to Claude Chat, 29% pointing to Claude Code, and 25% mentioning Claude App. Many users said in the community that even when no messages were actually sent, Claude encountered abnormal prompts such as showing that the message limit has been reached and the service is temporarily busy.
Earlier in the day, the number of related notifications from DownDetector quickly exceeded 7,000, then exceeded 15,000 within about 25 minutes, and later exceeded 20,000. Newsweek also quoted an Anthropic spokesperson as saying that services have been fully restored at Claude.ai, Cowork and Claude Code later.
The data obtained by CNBC is relatively conservative. The report pointed out that about 2,000 users on DownDetector were still reporting abnormalities at around 1:12 p.m. Eastern Time, which was significantly lower than the peak of about 6,000 in the morning, indicating that the service had indeed gradually stabilized before the official announcement of repairs.
It’s not just a one-day accident, Claude has actually experienced abnormalities many times in the past week
If the perspective is extended to the entire mid-April, this outage is not an isolated incident.Claude Status Historical records show that on April 13th, an exception occurred that affected the login of Claude.ai and Claude Code; on April 14th, usage and analytics admin API endpoint performance decreased; and there was even a wave of short-term outage related to Opus 4.6 on April 16th. In other words, Anthropic’s recent service stability has actually continued to be under pressure. The incident on April 15 was only the one that received the most attention and most directly impacted ordinary users.
This phenomenon also reflects the challenges faced by Claude’s infrastructure and service governance after its usage has expanded rapidly in recent months. Mashable mentioned in the report that Anthropic told the media that Claude’s new users once exceeded 1 million per day in March 2026. When Claude is no longer just a chat tool, but is gradually incorporated into enterprise APIs, AI coding workflows, research and collaboration scenarios, any abnormality in login, platform or API will be magnified from individual product problems to blockage of the entire work chain.
Claude Code is affected, and the impact is particularly obvious on developers
Compared with the brief unavailability of the pure chat service, another noteworthy point of this incident is that Claude Code was also affected. Anthropic’s official update mentioned that users who have logged in to Claude Code can still continue to use it, but the login mechanism was once invalid; Mashable and Newsweek also cited official statements to confirm that some users were unable to log in to Claude Code normally through Claude.ai. This means that for engineers who use Claude as their daily development assistant, the problem is not just “the chat window cannot be opened”, but the entire coding assistant workflow may be directly interrupted.
From an industrial perspective, this once again reminds the outside world that as generative AI tools become more and more deeply embedded in work scenarios, stability itself is already part of product competitiveness. No matter how powerful the model is, if the platform fails to log in, the API error rate soars, or the service is interrupted every three days, the actual trust cost for enterprises and professional users will continue to increase.
Key points to observe: What Anthropic needs to prove next is not just how fast it can be repaired, but whether it can cause fewer accidents.
This time, the Claude outage was finally repaired within a few hours, and Anthropic also continued to update the progress through the status page in a timely manner, which is considered qualified in terms of communication transparency. However, if we look back at the continuous abnormal records from April 13 to April 16, what the market will really care about is probably not the speed of a single repair, but whether Anthropic can return service stability to a more trustworthy level.
Especially when Claude, Claude Code and API are rapidly expanding the penetration of enterprises and developers, any interruption is not just as simple as “cannot chat today” for individual users, but may affect the research process, customer service process, internal tool chain and development rhythm. For Anthropic, the next test is not just to turn off this outage, but how to reduce the frequency of similar incidents happening repeatedly in a short period of time.
Source: KOCPC Chinese