Agent economics

Model the cost of multi-step AI agents.

Account for repeated system prompts, tool schemas, accumulated context, summaries, cache hits and billable tool calls.

Agent workflow

All token fields are per run unless stated otherwise.

Context-aware

System prompt optional estimator

Exact system tokens optional

Model

Service tier

Pricing date

Agent steps

Runs per month

New user tokens / step

Static tool-schema tokens

Tool-result tokens / step

Model output tokens / step

Context strategy

Summary-memory tokens

Billable web searches / run

Other priced tool calls / run

Cache the static system and tool-schema prefix after step one

Trace

History is re-sent in full-context mode. Fixed-memory mode substitutes the configured summary size.

Step	History tokens	Input tokens	Output tokens