Usage Tracking

Agents mode distributes work across multiple workers, which means token usage is also distributed. Multi gives you visibility into consumption at every level.

Viewing Usage

Click the usage icon at the top of the window to open the usage panel. In agents mode, this shows a comprehensive breakdown of token consumption.

Individual Subagent Usage

Each subagent’s usage is tracked independently. When viewing a specific subagent, the header displays:

Input tokens - Context sent to the subagent
Output tokens - Responses generated by the subagent
Total tokens - Combined consumption for that worker

This helps you understand which subagents are doing the heavy lifting and where tokens are being spent.

Combined Usage View

The main usage panel shows aggregate statistics:

Supervisor agent - Tokens used by the main coordinating agent
All subagents - Combined total across every spawned worker
Task total - Complete token consumption for the entire task

Understanding the Numbers

Why Agents Use More Tokens

Agent-based tasks typically consume more tokens than single-agent tasks because:

Coordination overhead - The supervisor spends tokens planning and delegating
Context duplication - Each subagent receives relevant context for its assignment
Reporting - Subagents summarize their work back to the supervisor

This is the tradeoff for being able to handle larger, more complex tasks.

Optimizing Usage

A few strategies to keep token consumption reasonable:

Strategy	Effect
Use agents only for complex tasks	Avoid coordination overhead on simple work
Clear, specific prompts	Reduces back-and-forth and clarification
Appropriate model selection	Match model capability to task complexity

Cost Visibility

The usage panel helps you make informed decisions about:

Whether a task justified agents mode
How to estimate costs for similar future tasks
Which types of work benefit most from delegation

Over time, you’ll develop intuition for when agents mode delivers value versus when it’s overkill.