[BUG] Pro plan: 4% of 5-hour limit consumed by ~5 text messages in claude.ai at midnight (off-peak)

Resolved 💬 2 comments Opened Mar 29, 2026 by aaharr Closed Apr 1, 2026

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?

Plan: Claude Pro
Time of occurrence: ~midnight PT (off-peak hours)
Interface: claude.ai web chat (not Claude Code CLI)
Windows app (alongside Arch WSL Claude Code)
Model: Sonnet 4.6
What happened:
A single claude.ai conversation consisting of approximately 5 text-only messages (no file uploads, no code execution, no agentic tasks) consumed ~4% of my 5-hour usage limit after a coding session with Claude Code, but also after limit reset. This would extrapolate to hitting the full limit after ~25 such messages, which is not consistent with normal usage expectations for a Pro plan, especially during off-peak hours.

I'm considering this separate issue from the reported claude code usage issues (which I am also experiencing) because of the off hours usage, in the Desktop UI running up without using claude code at the time (but logged in), and because these are simple text questions I'm asking in the UI and in Claude code, not complex tasks, not major tool usages.

What Should Happen?

Expected behavior: A short text Q&A conversation should consume a negligible fraction of a 5-hour session limit. I would expect 1% usage just due to rounding, but fractional at most.

Error Messages/Logs

Steps to Reproduce

Log into claude.ai on a Pro plan
Note current usage % in Settings → Usage (confirm starting from 0% or a known baseline)
Start a new conversation in the claude.ai web interface
Send approximately 5 text-only messages — no file uploads, no code execution, no Claude Code CLI, no agentic tasks. Conversational Q&A only.
Check usage % again after the conversation

Claude Model

Sonnet (default)

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.87

Platform

Anthropic API

Operating System

Other Linux

Terminal/Shell

WSL (Windows Subsystem for Linux)

Additional Information

<img width="886" height="735" alt="Image" src="https://github.com/user-attachments/assets/88818437-0daf-41cd-98eb-f3164959f628" />
After having the conversation needed to diagnose this issue, I went from 5% to 16% here. That's about 2% per simple text message at 2am pacific.

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗