[BUG] Claude Code usage limit hit after 7 minutes on new session

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?

I purchased a Claude Pro subscription 4 days ago with the goal of experimenting with Claude Code and building a better understanding of what it's capable of. I've used it briefly each day this week — the first couple of days I was testing basic coding tasks in an empty directory, yesterday I was investigating individual bugs in a real codebase. Today was my first attempt at something more ambitious: a parallel, agentic workflow.

The task was well-defined: analyze ~132 bug reports from a CSV file, investigate each one against the codebase, and write individual report files tagged by card ID. Claude Code handled it well initially — it read the file, identified the boundary row, created the output directory, and launched 5 parallel investigation agents covering ~100 bugs across different card groups. Seven minutes later, mid-task, I hit the usage limit and everything stopped.

This left a very bad taste in my mouth. I was optimistic about Claude Code as a tool I could rely on for serious development work. After this experience, it feels a lot weaker — primarily because it is less reliable. A tool that stops working mid-task, with no warning until it's already too late, is difficult to trust or plan around.
I thought that parallel agents are a Claude Code feature, not a hack. The task was the kind of agentic workflow the tool is designed for. But it turns out that feature is capable of consuming an entire 5-hour usage window in under 7 minutes on a Pro plan — and there's nothing in the onboarding or documentation that prepares you for that.

What Should Happen?

I'm not sure. I don't know what to expect, that's why I am learning and experimenting. I guess I expected to be able to use parallel agents for a task without immediately hitting the context threshold.

Error Messages/Logs

Steps to Reproduce

Try using Claude Code for investigation on a project with a sizable amount of code.

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.83 (Claude Code)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Windows Terminal

Additional Information

_No response_