[BUG] Claude Max weekly usage depleted unexpectedly fast for Claude Code subagent-heavy Gmail MCP sessions

Open 💬 2 comments Opened Jun 21, 2026 by DDDIGHE

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?

Summary

I am a Claude Max user. My weekly usage reached 100% after 6 Claude Code sessions, mostly scheduled briefing/email-harvest routines.

I understand that Opus, high effort, large context, MCP tools, and subagents consume significant quota. The problem is that the resulting Max weekly depletion seems difficult to reconcile with the visible workload, and Claude Code's /usage output does not provide enough account-level detail to understand whether this is expected.

Environment

  • Claude Code version: 2.1.185
  • Login method: Claude Max account
  • Model: Opus 4.8 with 1M context
  • MCP servers: Gmail MCP and Calendar MCP enabled
  • Effort/config: max effort, subagent-heavy workflow

Observed usage

Visible work this week was roughly:

  • 3 scheduled briefing/email-harvest routines
  • 2 manually opened Claude Code sessions
  • 1 client-side conversation with only one question

Local de-duplicated Claude Code usage for the 3 briefing routines was approximately:

  • 95 unique model requests
  • 131K input tokens
  • 1.8M cache creation tokens
  • 16.0M cache read tokens
  • 249K output tokens

/usage attributed much of the usage to:

  • 100% subagent-heavy sessions
  • /email-harvest
  • Gmail MCP server usage
  • >150k context sessions

What Should Happen?

Could Claude Code provide clearer diagnostics for Max usage depletion, especially for:

  1. Subagent-heavy sessions
  2. MCP tool result contribution
  3. Cache creation/cache read contribution
  4. Long-context contribution
  5. Difference between local /usage approximation and actual account-level Max usage

Error Messages/Logs

Steps to Reproduce

none

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.185

Platform

Other

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_

View original on GitHub ↗

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