Tool Visualization

Overview

One of C3's most powerful features is the ability to see exactly what the agent is doing. Every tool execution — file reads, writes, command runs, and more — is displayed as a visual card with full details.

This transparency lets you:

• Understand what changes the agent is making
• Catch potential issues before they happen
• Learn from the agent's problem-solving approach
• Debug unexpected behavior

Tool Cards

Each tool execution appears as a card in the chat. Cards show:

Header

• Tool Icon — Visual indicator of the tool type
• Tool Name — The specific tool being used
• Status — Pending, running, complete, or error
• Expand/Collapse — Toggle to show/hide details

Input Section

Shows the parameters the agent passed to the tool:

{
  "file_path": "/src/components/Button.tsx",
  "old_string": "const Button = () => {",
  "new_string": "const Button: React.FC<ButtonProps> = () => {"
}

Output Section

Shows the result of the tool execution. For file reads, this is the file content. For commands, this is stdout/stderr. Large outputs are truncated with an option to expand.

Common Tools

Here are the tools you'll see most frequently:

File Operations

Command Execution

Special Tools

Approval Flow

When a tool requires approval (based on your permission settings), the card shows an approval prompt:

Pending Approval State

• Card is highlighted with an attention indicator
• Shows what the agent wants to do
• Provides Approve and Reject buttons
• You can expand to see full details before deciding

Making Decisions

• Approve — Execute the tool and continue
• Reject — Skip the tool and let the agent know
• Approve All — Approve all similar requests in the future
• Tell the Agent What to Do — Provide specific instructions for the agent to follow

Allowed Tools

You can configure certain tool patterns to auto-approve. For example, you might allow all Read operations but require approval for Write operations. See the Permission System documentation for details.