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.