[BUG] Severe session limit reduction for Pro users (Hitting 100% in ~3 simple prompts)
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
Hi Claude Code team! First of all, I want to say that I absolutely love this product. It has become an essential tool in my daily development workflow.
However, I am writing to express a growing concern that many users, including myself, are currently experiencing regarding session usage limits.
The Issue:
Recently, the usage limit per session seems to have been drastically reduced. As a Pro subscriber who has been using the tool for a while, I've never seen my limits drain this fast. Right now, I am hitting 100% usage after just about 3 simple prompts. This effectively makes the new session limit unusable for standard development tasks.
Context & Impact:
This is happening on the Pro tier.
It seems to be affecting a significant number of users, though perhaps not everyone.
Because there hasn't been any official communication or acknowledgment yet, it feels a bit like user reports might be misunderstood as exaggerations without proper investigation.
Request:
I genuinely hope this is an unexpected bug rather than an intended behavior. Could the engineering team please give this the priority it deserves and look into it? If this is an intended change to the session limits, it would be highly appreciated if you could release an official statement explaining the new thresholds so we can adjust our expectations and usage.
Thanks for your hard work and for building such a great tool!
What Should Happen?
As a Pro subscriber, a standard session should easily accommodate a prolonged development workflow. Sending ~3 simple, low-context prompts should only consume a very small fraction of the session limit. The usage percentage should accurately reflect the actual token consumption or context size, allowing for continuous, multi-turn interactions without hitting the 100% cap almost immediately.
Error Messages/Logs
Steps to Reproduce
Start a fresh, new session in Claude Code using a Pro account.
Submit a very simple, low-context prompt (e.g., "Write a hello world in Python" or a basic refactoring question).
Monitor the session usage indicator.
Submit 1 or 2 more simple follow-up prompts.
Observe that the session usage rapidly spikes to 100%, effectively blocking further interaction in a fraction of the usual time.
Claude Model
None
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
2.1.81 (Claude Code)
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
iTerm2
Additional Information
_No response_
Showing cached comments. Read the full discussion on GitHub ↗
5 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
Same issue I am facing
I have do only /compact. This is 20percent of hourlimit and 2Percent Weekly Limit. I cant do work with it,
I thought I am the only one getting this. I started question my prompting skills. Was about to take prompt engineering course again. Sometimes two to three or sometimes only single prompt is enough to hit the 4hrs session limit.
I've been trying to work around this issue as well. It's gotten worse and worse as my project has gotten bigger. I've broken large files into smaller modules. I have tried having it spawn agents to take tasks out of the main context. I have updated Claude.md. I've updated memory.md, and nothing seems to be improving the situation. I dug into it earlier today and burned up the last roughly 25% of my week trying to solve it.
Apparently, part of the issue is that even though there are instructions, guidelines, and rules written into Claude.md, it just ignores them. It also doesn't prune Claude.md like it's supposed to, so my file had grown to over 400 lines that were loaded and re-read into every fresh conversation and at each turn of the conversation. Basically, just opening up a window and asking it anything was burning 8 to 12% of a session which compounded from there with every ask.
I came across a solution in some of the other reports that I'm going to try when my weekly limit resets on Thursday now to create hooks, pre and post-tool hooks that will supposedly prevent it from doing things it's not supposed to be doing which is what I thought Claude.md was for, but apparently there's another layer of rules that basically physically prevent it from going off and doing its own thing, so I'm going to try that and see how that works. It'll probably only eat a quarter of next week's tokens to build it and test it....
It's feeling like this is becoming less of a tool and more of a burden with every weekly limit that gets hit faster and faster. They need to fix the situation, or they're going to start losing customers as quickly as they got them.