Feature Request: Show baseline context separately from conversation context in status line
Resolved 💬 5 comments Opened Dec 11, 2025 by jeansebastienvincent Closed Feb 11, 2026
Summary
When using /clear, the context doesn't go to zero because there's significant baseline context that loads with every conversation. It would be very helpful to see this baseline context displayed separately from conversation context in the status line.
Current Behavior
The status line receives only aggregate context data:
{
"context_window": {
"total_input_tokens": 15234,
"total_output_tokens": 4521,
"context_window_size": 200000
}
}
After /clear, users see a non-zero context count with no visibility into what's consuming that space.
Proposed Enhancement
Provide a breakdown of context usage by source in the status line JSON:
{
"context_window": {
"total_input_tokens": 15234,
"total_output_tokens": 4521,
"context_window_size": 200000,
"breakdown": {
"system_prompt": 2500,
"claude_md_files": 4200,
"mcp_server_definitions": 8000,
"tool_definitions": 3500,
"conversation": 1034
}
}
}
Use Cases
- Understanding overhead - Users with large CLAUDE.md files or many MCP servers can understand their "cost" before starting work
- Optimizing configuration - Helps identify if CLAUDE.md or MCP configs should be trimmed
- Context management - Better decisions about when to
/clearor/compact - Debugging - Understanding why context is high after clearing
Additional Suggestions
- A
/context-breakdowncommand to show detailed context composition on demand - Display baseline vs conversation in the default status line format
Environment
- Claude Code CLI
- Users with multiple CLAUDE.md files and MCP servers configured
This issue has 5 comments on GitHub. Read the full discussion on GitHub ↗