Feature Request: Compact/Agent Mode - Hide Tool Execution Output

Context

With the advent of agents and skills, developers are increasingly using Claude Code as an orchestration layer rather than a direct coding assistant. In these workflows, users care about what the agent accomplished, not every bash command or file read that happened along the way.

Currently, Claude Code displays:

1. The raw tool execution output (bash commands, file contents, etc.)
2. Claude's formatted summary of that same information

This creates two problems:

• Context bloat: The same information consumes tokens twice
• Noise: When orchestrating complex multi-step tasks, the terminal becomes a wall of intermediate outputs that obscure the actual results

Proposed Solution

Add a "compact mode" or "agent mode" setting that:

• Hides raw tool output (bash commands, file reads, grep results, etc.)
• Shows only Claude's processed response (the summary, analysis, or result)
• Optionally shows a collapsed indicator (e.g., [3 tools executed] that can be expanded if needed)

Configuration Options

{
  "outputMode": "compact"  // or "verbose" (current default)
}

Or a CLI flag:

claude --compact "your task"

Or an interactive toggle:

/compact on
/compact off

Use Cases

1. Agent orchestration: Running complex multi-agent workflows where only final results matter
2. Skill execution: Using skills that perform many internal operations
3. Demo/presentation: Showing Claude Code to stakeholders without implementation noise
4. Context preservation: Reducing token usage in long sessions

Acknowledgment

There are valid reasons to see tool output (debugging, transparency, learning). This isn't about removing that capability - it's about giving users the choice to hide it when they don't need it.

A toggle between modes would serve both use cases:

• Verbose mode (current): Full transparency, see everything
• Compact mode (new): Results-focused, agent-style output

Related

This becomes increasingly important as Claude Code moves toward more autonomous agent workflows where dozens of tool calls may happen for a single user request.