AgentScout / Free tools / AI Agent Run Cost Estimator

AI agent run cost estimator

Work out what one multi-step agent task really costs, with a growing context, tool results, prompt caching and retries, then compare it on 33 GPT, Claude, Gemini and Grok models.

Free, no sign-up. Runs entirely in your browser: nothing you type is sent to AgentScout or any AI provider.

Model

Steps (model calls) per task

System prompt + tool definitions (tokens)

User request + starting context (tokens)

Tool result added per step (tokens)

Model output per step (tokens)

Cached share of repeated input (%)

Extra steps from retries (%)

Tasks per day

The same agent on every model

Sorted by cost per task. Click a row to select it.

Model	Provider	Per task	Per month

How it works

An agent re-sends its whole conversation on every step, so input grows: step k sends the system prompt and tools, the starting context, and every earlier tool result and model reply. The estimator adds those up step by step. The cached share applies to input that was already sent on an earlier step (prompt caching), billed at the cached rate; the first step is always full price. Retries add extra steps at the average step cost.

Prices are the same verified list the AI API cost calculator uses, checked on 2026-10-19 from each provider's official pricing page (standard tier, short-context rates). This tool does not list context-window sizes: check the provider page to make sure your final-step context fits the model.

FAQ

Why is an agent task so much more expensive than one chat request?

Each step re-sends the full conversation so far, including every tool result, so input tokens grow with every step. A 10-step task can use many times the tokens of a single prompt.

How does prompt caching change the cost?

Input already sent on an earlier step can often be served from the provider's prompt cache at a lower cached rate. Set the cached share to how much of that repeated input you expect to hit the cache; 0% means no caching.

Are reasoning tokens included?

Only if you add them. Thinking tokens are billed as output, so raise the output-per-step figure if your model reasons before acting.

Where do the prices come from?

The same list as the AgentScout cost calculator, copied from the official OpenAI, Anthropic, Google and xAI pricing pages on 2026-10-19. Confirm on the provider page before you budget.

More free AI tools built by : AI API Cost Calculator · AI Token Counter · AI Prompt Builder · AI Agent Setup Generator · MCP Config Builder · llms.txt Generator · AI Tool Schema Builder · AGENTS.md & Rules File Generator · AI Agent Stack Builder · System Prompt Linter · Context Window Planner · Open-Source License Checker for AI Tools · Build vs Buy AI Agent Calculator · MCP Server Picker · JSON Output Schema Validator · AI Tool Comparison Card Maker · AI Agent Readiness Checker · Free AI image generator · All free tools