Using sub-agents to cut AI costs in practice

This post shares tips on reducing API costs when using AI agents by delegating simpler tasks to cheaper sub-agents. Instead of routing everything through an expensive main agent, low-complexity work goes to a smaller, cheaper model.

Running a capable AI agent like hermes-agent can get expensive fast, because every task runs through a powerful (and costly) model. The approach here is to set up sub-agents — smaller AI models that handle specific, routine tasks — so the main agent only handles decisions that really need it.

For example, file reading, simple searches, or text formatting can be handled by a lighter model at a fraction of the cost. For regular hermes-agent users, this pattern is a practical way to keep usage affordable without sacrificing quality on complex tasks.

Key points

  • Route only high-level decisions to the main (expensive) agent
  • Assign repetitive or simple tasks to cheaper sub-agents
  • This can significantly lower total API costs
  • File handling, search, and formatting are good candidates for sub-agents
  • Applies directly to hermes-agent setups that use multiple model tiers

Quick term guide

API costs
Fees paid when software calls an online service programmatically.
AI agents
AI agents are AI tools that can carry out steps toward a goal, not just answer once.
AI agent
An AI program that can inspect information and suggest what to do next.
sub-agent
A smaller AI agent that handles one specific part of a larger task, running separately from the main agent.
routing
Automatically deciding which AI model handles a request based on how complex or simple it looks.
hermes-agent
A likely name for Nous Research’s agent-style AI tool or service.
AI models
The core brain or underlying program that powers an artificial intelligence tool.
AI model
A program that can understand prompts and produce text, code, or answers.
Read original