[BUG] Claude Max 5x: 1M context window auto-triggered in Cowork session, deducting usage credits without user request

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?

While using Claude Sonnet in a Claude Cowork session on claude.ai (not standard chat,
not Claude Code CLI), the system silently routed my conversation through the 1M context
window and deducted €22.27 from my pre-loaded usage credit balance, 223% of my €10
monthly spend limit. I never requested extended context. Weekly plan usage was at 2-3%
at the time.

Actual behavior:

• €22.27 deducted from pre-loaded credit balance (223% of €10 monthly spend limit)
• A "Need more usage?" modal appeared post-hoc; the deduction had already occurred
• Conversation became permanently unusable; "Try again" had no effect
• Automated support (Fin) became unresponsive, Conversation ID: 215474592791250

What Should Happen?

• claude.ai Cowork should never trigger 1M context routing without explicit user opt-in
• Usage credits should not be deducted for a context window the user did not request
• The €10 monthly spend limit should function as a hard cap
• Weekly plan usage at 2-3% confirms this was not a plan exhaustion issue

Error Messages/Logs

API Error: Usage credits required for 1M context · turn on usage credits at
claude.ai/settings/usage, or use --model to switch to standard context

Steps to Reproduce

1. Start a long multi-turn conversation in a Claude Cowork session on claude.ai
2. System silently routes to 1M context without user confirmation
3. €22.27 is deducted from usage credits (223% of €10/month spend limit)
4. Conversation becomes blocked with the error above
5. "Try again" does not resolve it; the conversation is permanently broken

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

N/A - bug is in claude.ai Cowork web UI, not the CLI

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Other

Additional Information

Note: the error references --model, a Claude Code CLI-only flag, in a web UI context.
This suggests the error text is shared between CLI and web without adaptation; a web
user has no way to pass --model.

Screenshots attached: Usage settings showing 2-3% weekly plan usage; €22.27
spent / 223% of €10 limit with credits toggle enabled; post-hoc "Need more usage?"
modal; error screen.

Related issues: #65514, #46780, #45497

<img width="680" height="431" alt="Image" src="https://github.com/user-attachments/assets/9e3b5a76-f940-4563-8f18-a9d92a4a31f5" />
<img width="975" height="735" alt="Image" src="https://github.com/user-attachments/assets/c3b4902d-72df-40db-8faf-6a4f59c39767" />
<img width="1005" height="739" alt="Image" src="https://github.com/user-attachments/assets/85640e6d-0fa5-4d81-9cd0-fcb75e25f2e7" />
<img width="641" height="449" alt="Image" src="https://github.com/user-attachments/assets/5e37fd76-7b82-4577-af8d-c07a008d2e67" />