Agent Workflow Cost Calculator

Why model your full agent pipeline?

Single-call calculators miss the real picture. A production AI agent typically runs 5–10 LLM calls per user request — routing, retrieval, reasoning, validation, and response generation. Each step has different token counts, different models, and different call frequencies.

The Agent Workflow Calculator lets you model every step independently so you can see exactly where your AI budget is going before you deploy — and identify which swap would save the most money.

Common multi-step agent patterns

• •RAG pipeline: Intent router → vector search → context expansion → answer generation
• •Tool-use agent: Planner → tool selector → executor → summarizer
• •Reasoning chain: Draft → critic → revise → final
• •Classification + generation: Cheap classifier first, expensive generator only when needed

How to reduce multi-step pipeline costs

• •Route cheap tasks to Haiku, Flash, or GPT-4o mini — save 80–95% vs flagship
• •Short-circuit early: classify intent first, skip expensive steps for simple queries
• •Cache repeated context (system prompts, retrieved docs) across turns
• •Cap output tokens per step — unnecessary verbosity is expensive

How this calculator works

Unlike a single shared request count, each step here has its own "runs per day" — so a retrieval step that fires on every message and a validation step that only fires occasionally can be modeled accurately in the same workflow. Every step's cost per run is its input tokens divided by one million times the model's input price, plus output tokens divided by one million times the output price, using the current prices shown in the calculator; multiplying by that step's runs per day and by 30 gives its monthly cost.

The calculator also computes what percentage of your total spend each step represents, and — separately from the same-provider swaps other calculators suggest — scans every model across every provider for a cheaper alternative for each step, so you can see the single biggest cost driver in your pipeline and the largest available saving.

How to use it

1. 1.Add one step per distinct model call in your pipeline, using names that match your real architecture (router, retriever, generator, etc.).
2. 2.Set each step's model and its own input/output token counts — these usually differ a lot between steps.
3. 3.Set each step's runs per day independently; a step that only triggers on certain requests should have a lower number than one that runs every time.
4. 4.Check the "most expensive step" and percent-of-total breakdown to see where your budget is actually going.
5. 5.Review the cross-provider scenario suggestions for the steps with the highest cost.

Common mistakes

• ⚠Giving every step the same run frequency — conditional steps (retries, escalations, fallbacks) run less often than the main path — entering the same number for all of them overstates their real cost.
• ⚠Optimizing the cheapest step first — the percent-of-total breakdown exists so you fix the step that actually moves your bill, not the one that's easiest to change.
• ⚠Ignoring cross-provider suggestions — same-provider swaps are convenient, but the biggest savings often come from switching providers entirely for a specific step — check the scenario comparisons.
• ⚠Not re-checking after a swap — once you change a step's model, its cost-per-run and percent-of-total change too — re-run the calculation to see the new bottleneck.

Frequently asked questions