Anthropic has reset five-hour and weekly limits for all Claude Code Pro and Max subscribers after a software bug caused some sessions to burn through quotas far faster than intended. The company says the issue has been fixed, and the move is meant to compensate users whose available work time was eaten up by the fault rather than by real usage.
The bug affected Claude Code usage limits by spinning up too many parallel sub-agents and tool calls, which meant the service did a lot more work than it should have. For developers, that translated into daily limits disappearing after a handful of ordinary prompts – exactly the kind of quiet failure that turns a coding assistant into a paperweight.
What went wrong in Claude Code
According to Anthropic, the problem was not a new policy on usage caps and not a change in Dynamic Workflows, despite speculation in developer communities. The company said the fault sat inside how Claude Opus 4.8 handled requests, where the system began triggering more simultaneous tool invocations than the original design allowed. In other words: the model was overworking itself, and users paid the bill.
That explanation matters because quota systems are supposed to be boring. When they fail, trust goes with them, especially in paid developer tools where every request is part productivity, part metering anxiety. Anthropic is also playing catch-up against rivals like OpenAI and Google, which have been under pressure to make coding products feel less fragile and less stingy.
Why the Claude Code reset affects paid users
The reset applies to all Pro and Max subscribers, which makes this more than a courtesy gesture. It is Anthropic acknowledging that the service consumed a scarce resource – user quota – because of its own error. That is the right call, and also the bare minimum if the company wants developers to keep treating Claude Code as a serious working tool rather than a clever demo that occasionally eats its own lunch.
- Five-hour limits were reset for affected Pro and Max users
- Weekly limits were reset too
- Anthropic says the bug has already been fixed
- The company blames excessive parallel tool calls, not Dynamic Workflows changes
What Claude Code users will watch next
The bigger question is whether this turns into a one-off embarrassment or a sign that Claude Code’s orchestration layer is still too eager to scale up work behind the scenes. If Anthropic wants to keep winning developers who live by quotas, logs, and predictable output, it will need to prove that its fix is durable – because nothing kills enthusiasm for an AI coding assistant faster than wondering whether your next prompt will vanish into a quota black hole.