Claude Code weekly limits: September 14 permanent 25% hike is a real-world cut

What was announced

Anthropic's developer tools account, ClaudeDevs, said on August 29 that it will permanently raise weekly usage limits for its coding agent, Claude Code, by 25% over the original baseline starting September 14. The change applies to Pro, Max, Team, and seat-based Enterprise plans. Anthropic said the current temporary 50% boost will remain in effect until September 14.

Anthropic official website

Techmeme summarized the announcement by noting that while weekly limits will permanently rise 25% for most plans starting September 14, that actually represents a 17% cut compared to the current 50% boost. Do the math, and once the temporary measure is lifted, the limit users actually feel will end up lower than what they have now.

What's behind this

This move didn't come out of nowhere. Claude Code had already extended its temporary 50% weekly limit increase through August 31, back on August 18. At the time, Anthropic said demand for the model was strong enough that capacity could stay tight for several weeks, and it signaled it was considering turning that temporary boost into permanent policy. This latest announcement is effectively the conclusion of that review — notably, the permanent increase ended up smaller than the temporary one it replaces.

In a follow-up report on August 30, The Decoder said Anthropic isn't just adjusting the numbers — it's also working on changes that give users more control and transparency over their usage, aiming to make people feel like they're actually getting more out of Claude. Anthropic hasn't said yet which features will roll out or when.

Claude Code is an agentic tool that reads, fixes, and even runs code right in the terminal, so the amount of tokens a user can burn through in a week effectively caps how much work they can get done. For Anthropic, the pattern seems to be: loosen limits temporarily when demand spikes, then dial them back to a sustainable level once things stabilize. Anthropic is the company behind Claude — Amazon and Google are investors and cloud providers, not partners in Claude's development. This limit policy, too, was decided and announced by Anthropic alone.

A recently added automatic feedback tool fits into this same push for improvement. According to Anthropic's official documentation, starting with Claude Code version 2.1.238, if a tool or command keeps failing — or if a user flags an error — Claude can draft feedback locally. That draft only gets sent after the user reviews it, users can choose whether to include conversation history, and the feature can be turned off in settings entirely. It doesn't affect weekly limits on its own, but it's another sign that Anthropic is trying to make both usage visibility and its product development process more transparent.

What this means going forward

If you're currently using Claude Code, it's worth knowing ahead of time that the limit you actually feel could shrink once September 14 hits. Teams doing heavy coding work in particular should take a look at their usage patterns around that date. This pricing structure applies the same way to development teams using Claude Code anywhere, including in Korea, so if your project runs high usage volumes, it's safer to adjust your work schedule around the September 14 transition.