Subagents can raise token cost when they repeat research
use of Claude Code and Cursor showed that do not always cut costs. Simple efficiency ideas can work, such as using cheaper models for easier tasks and shrinking before it grows too large.
The problem was that some subagent setups created duplicate work. Before coding, the main agent sent two to study the , but after they returned their findings, the main agent read the same files again.
Some rereading makes sense for code edits because changes must land in the right place, but the rereading spread to related too. The result was about 3 extra minutes of waiting, with the using about 200,000 tokens to save only about 20,000 tokens of extra reading by the main agent.
Key points
- Two first ed the .
- The main agent then reread the same files after receiving their findings.
- The used about 200,000 tokens.
- The saved main-agent reading was only about 20,000 tokens.
- can increase cost if their is duplicated.