[Bug] Instant Auto-Compact increases token consumption even when disabled

Open 💬 4 comments Opened Dec 10, 2025 by BnnQ

Bug Description
after 2.0.64 update (probably because of instant auto-compact introduced) token usage spiked crazy, even though I have auto-compaction disabled. Previously it was only consuming tokens during compaction itself (and much less), but now even without doing anything with the compaction (explicitly), "background" token consumption during claude work is much higher (while running compaction itself no longer consumes tokens, but still). And the more the context is filled, the more tokens are consumed, sometimes I notice how 10% of the 5-hour limit is suddenly eaten up while claude was just doing its regular work at ~180K context. Next message? Another 5-10% eaten. And it definitely wasn't the case previously, it was absolutely possible that 5-10% of 5-hour limit would be eaten within one "message", but progressively, not all of a sudden.

Environment Info

  • Platform: win32
  • Terminal: windows-terminal
  • Version: 2.0.64
  • Feedback ID: 74574dbe-baa8-4c9b-a359-9bf273d4c2b0

Errors

[{"error":"MaxFileReadTokenExceededError: File content (43718 tokens) exceeds maximum allowed tokens (25000). Please use offset and limit parameters to read specific portions of the file, or use the GrepTool to search for specific content.\n    at xtL (B:/~BUN/root/claude.exe:1252:10587)\n    at async call (B:/~BUN/root/claude.exe:1261:972)\n    at async <anonymous> (B:/~BUN/root/claude.exe:1478:1511)\n    at processTicksAndRejections (native:7:39)","timestamp":"2025-12-10T14:41:15.435Z"},{"error":"MaxFileReadTokenExceededError: File content (43718 tokens) exceeds maximum allowed tokens (25000). Please use offset and limit parameters to read specific portions of the file, or use the GrepTool to search for specific content.\n    at xtL (B:/~BUN/root/claude.exe:1252:10587)\n    at async call (B:/~BUN/root/claude.exe:1261:972)\n    at async <anonymous> (B:/~BUN/root/claude.exe:1478:1511)\n    at processTicksAndRejections (native:7:39)","timestamp":"2025-12-10T14:41:42.096Z"},{"error":"MaxFileReadTokenExceededError: File content (43718 tokens) exceeds maximum allowed tokens (25000). Please use offset and limit parameters to read specific portions of the file, or use the GrepTool to search for specific content.\n    at xtL (B:/~BUN/root/claude.exe:1252:10587)\n    at async call (B:/~BUN/root/claude.exe:1261:972)\n    at async <anonymous> (B:/~BUN/root/claude.exe:1478:1511)\n    at processTicksAndRejections (native:7:39)","timestamp":"2025-12-10T14:43:05.797Z"},{"error":"MaxFileReadTokenExceededError: File content (43718 tokens) exceeds maximum allowed tokens (25000). Please use offset and limit parameters to read specific portions of the file, or use the GrepTool to search for specific content.\n    at xtL (B:/~BUN/root/claude.exe:1252:10587)\n    at async call (B:/~BUN/root/claude.exe:1261:972)\n    at async <anonymous> (B:/~BUN/root/claude.exe:1478:1511)\n    at processTicksAndRejections (native:7:39)","timestamp":"2025-12-10T14:44:36.726Z"},{"error":"MaxFileReadTokenExceededError: File content (43718 tokens) exceeds maximum allowed tokens (25000). Please use offset and limit parameters to read specific portions of the file, or use the GrepTool to search for specific content.\n    at xtL (B:/~BUN/root/claude.exe:1252:10587)\n    at async call (B:/~BUN/root/claude.exe:1261:972)\n    at async <anonymous> (B:/~BUN/root/claude.exe:1478:1511)\n    at processTicksAndRejections (native:7:39)","timestamp":"2025-12-10T14:44:55.590Z"},{"error":"Error: Request was aborted.\n    at makeRequest (B:/~BUN/root/claude.exe:1096:3939)\n    at processTicksAndRejections (native:7:39)","timestamp":"2025-12-10T15:29:51.586Z"},{"error":"Error: Request was aborted.\n    at _createMessage (B:/~BUN/root/claude.exe:127:3151)\n    at processTicksAndRejections (native:7:39)","timestamp":"2025-12-10T15:44:22.625Z"}]

View original on GitHub ↗

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