Budget & Cost Tracking

Agent HQ tracks the cost of every AI interaction, giving you visibility into spending.

How costs are calculated

Every message and task execution consumes tokens — the units AI models use to process text. Costs are calculated as:

cost = (input_tokens × input_price) + (output_tokens × output_price)

Different models have different pricing. More capable models (like Claude Opus 4.6 or GPT-4o) cost more per token than lighter models (like Claude Haiku 4.5 or Gemini Flash).

Per-task cost tracking

Each task shows its cost:

• Input tokens — tokens sent to the model (your instructions + code context)
• Output tokens — tokens generated by the model (code + explanations)
• Cost — the total dollar cost of the task

This information is available in the task detail view.

Per-message tracking

Individual chat messages also track token usage, so you can see the cost of each conversation turn.

Cost optimization tips

1. Use lighter models for simple tasks — Haiku or Gemini Flash for quick fixes
2. Write clear instructions — reduces back-and-forth and wasted tokens
3. Set project instructions — gives Pilot context upfront, reducing the need for large code reads
4. Use Cloudflare Workers AI — free models for experimentation
5. Review before retrying — understand why a task failed before running it again