[BUG] 10-12% usage spike when switching accounts and continuing task

Resolved 💬 4 comments Opened Jan 27, 2026 by scarabdesign Closed Mar 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?

When switching between logged-in accounts in Claude Code and continuing work on the same task/project, there is an immediate 10-12% usage spike on the newly logged-in account, even though the context was already built and paid for by the first account.

Actual Behavior

Account B immediately shows 10-12% usage spike, suggesting the entire project context is being reloaded and billed to the new account, even though this context was already built and paid for by Account A.

Why This Is Problematic

  • Both accounts are being charged for the same context loading
  • This feels like double-billing for the same work
  • Makes multi-account workflows inefficient and expensive
  • Suggests poor account isolation or inefficient context management

What Should Happen?

Account B should only consume tokens for new work and responses, not for re-loading project context that was already built by Account A. The context loading should be minimal or on-demand.

Error Messages/Logs

Steps to Reproduce

  1. Log in to Claude Code with Account A
  2. Start working on a task/project, building up context
  3. Log out and switch to Account B using claude logout then claude login
  4. Continue working on the same task/project
  5. Check usage with /usage or /context
  6. Observe immediate 10-12% usage consumption on Account B

Claude Model

Opus

Is this a regression?

No, this never worked

Last Working Version

_No response_

Claude Code Version

2.1.20

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

PowerShell

Additional Information

_No response_

View original on GitHub ↗

This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗