[BUG] Context limit reached in under 1 minute on fresh conversation (Max Plan + Extra Usage)

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?

Environment:
Claude Code version: 2.1.29
Plan: Claude Max
Extra usage: Enabled (20€ purchased)
Model: Sonnet 4.5 (claude-sonnet-4-5-20250929)
OS: Ubuntu (Linux)
Working directory: Local project (~standard Next.js web app)

Context is exhausted in a single interaction
Error message: "Context limit reached · /compact or /clear to continue"
Compacting fails: "Error compacting conversation"
Status shows: "Context left until auto-compact: 0%"

What Should Happen?

A fresh conversation should allow multiple interactions before hitting context limits
With Max plan + extra usage enabled, I expected more capacity
The compact feature should work to recover the conversation

Error Messages/Logs

Fresh conversations hit "Context limit reached" within 30-60 seconds of starting, even with a brand new session. The auto-compact feature fails with "Error compacting conversation" and shows "Context left until auto-compact: 0%".

Steps to Reproduce

Open Claude Code with Max plan subscription
Start a fresh conversation (click + or Ctrl+T)
Send a single prompt asking to read a project markdown file (~2-3KB) and fix a bug, attaching one screenshot
Claude reads 4-5 files and immediately hits context limit
/compact fails, conversation becomes unusable

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.29

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

Confusion with Extra Usage
I purchased €20 of extra usage thinking it would help with this issue. Later I learned that:

Extra usage = extends weekly usage quota (tokens over time)
Context limit = fixed 200K token window per conversation (cannot be extended)

This distinction is not clear in the UI or documentation. When hitting limits so quickly, users naturally assume purchasing "extra usage" will help.

What I was trying to do
Simple prompt: "Read the CLAUDE.md file in the project root and fix a bug where suppliers aren't loading when scanning a delivery note" + attached 1 screenshot.
Claude searched 5 patterns, read 4 files, then immediately hit the limit before providing any solution.
Impact

Cannot effectively use Claude Code for development work
Sessions last less than 1 minute
Paid for Max plan + extra usage but experience is severely degraded
This is blocking my workflow completely

Possible Related Issues

#9094 - Unexpected change in Claude usage limits
#8651 - Unexpected Usage Depletion on MAX20 Plan
#12450 - Extra usage enabled but issues persist

Suggested Improvements

Clearer distinction in UI between "usage limits" vs "context limits"
Better error messages explaining WHY context filled so fast (show token breakdown)
More aggressive automatic compacting before hitting 100%
Option to exclude certain files from context or set max file read size