Anthropic quietly throttles Claude Code limits for heavy users

Anthropic just gave its most lucrative power users a brutal reality check, and it forgot to send the memo first.

Starting Monday morning, heavy users of Anthropic's Claude Code began hitting unexpectedly harsh usage limits. The sudden throttle mostly slammed subscribers on the premium $200-per-month Max plan, leaving them stranded mid-workflow.

Silence from the top

Instead of shipping code, developers found themselves staring at a blunt message: Claude usage limit reached. It offered only a countdown timer, usually a few hours long, until the quota reset.

Because Anthropic made zero public announcements about shifting policy, the project's GitHub issues quickly turned into a sounding board for confused subscribers. Many assumed their accounts had suffered a glitch or an accidental downgrade. One frustrated user pointed out that firing off a few requests in 30 minutes could not realistically burn through a supposed 900-message ceiling.

When asked for comment, Anthropic confirmed the issues to TechCrunch but ducked the tough questions, claiming only that it was aware of slower response times for some users and working on fixes.

To make matters weirder, the network hiccups extended well beyond Claude Code. API developers reported repeated overload errors across the same stretch, with the company's status page tracking six separate incidents over four days. Yet, in typical enterprise fashion, the uptime dashboard stubbornly reported a spotless 100% for the week.

The $1,000-a-day loophole

Part of the mess traces right back to how Anthropic structures its pricing.

The $200 Max plan promises 20 times the usage of a Pro tier, which itself offers five times the capacity of a free account. But the baseline free tier is explicitly defined as varying by demand with no hard ceiling attached. When you multiply a moving target, nobody knows where the fence actually is.

At the same time, anyone doing the math saw this coming. One developer noted that under the $200 monthly plan, heavy Claude Code sessions regularly racked up the equivalent of over $1,000 in raw API tokens in a single day.

Anthropic was effectively subsidizing an unsustainable firehose of compute. But while power users understood the economic math, the covert cutback left a bitter taste. As one user put it, the radio silence is what destroys trust.

Why it matters

We think Anthropic made the right economic call but handled the rollout about as poorly as possible.

For professional developers relying on Claude Code for daily work, unpredictability is fatal. You simply cannot build critical pipelines around an agent that might shut off without warning. Competitors like Gemini or Kimi are waiting in the wings, but as affected users openly admitted, none of them match Claude's nuanced coding skill right now. That moat buys Anthropic leverage, but it does not buy unlimited patience.

Across the broader landscape, this clash proves that flat-rate monthly subscriptions are failing the agentic coding era. As coding tools chain multiple autonomous reasoning loops together, token burn skyrockets. Forcing hard caps is Anthropic's only defense against massive server deficits, but implementing stealth throttling without clear meter dashboards signals that consumer-style billing cannot survive enterprise workloads.