Do function definitions really cost tokens?

Yes. The tools or functions you pass are serialized and prepended to your prompt as input tokens on every request. A handful of verbose schemas can add hundreds of input tokens to each call.

Are the definitions billed on every request?

Yes — unless you use prompt caching for the stable tools block. Otherwise the full schema is re-sent and re-billed each call, which is exactly the recurring overhead this tool quantifies.

How are the schema tokens estimated?

The tool counts the serialized JSON using a roughly four-characters-per-token heuristic, plus a small per-function structural overhead the API adds when formatting tools. It is an estimate, usually within ~10% for typical schemas.

How do I cut function-calling overhead?

Trim verbose descriptions, send only the tools relevant to each request, shorten parameter names, and enable prompt caching on the tools block so the stable schema is not re-billed every call.

Is my schema uploaded?

No. Parsing and counting happen entirely in your browser. Nothing you paste is sent anywhere.

Function Calling Token Cost Calculator

Function calling token cost calculator

Function (tool) calling is one of the most underestimated cost lines in an LLM app. The JSON schemas you pass — names, parameter types, and especially long natural-language descriptions — are serialized and prepended to your prompt as input tokens on every request. A few rich tools can quietly add hundreds of input tokens per call. This tool measures that hidden overhead and projects its monthly cost.

How it works

Paste the tools or functions array you send with each request. The tool serializes and counts it using a four-characters-per-token heuristic plus a small per-function structural allowance for the wrapping the API adds. Because definitions are re-sent on every call (unless cached), the per-request overhead multiplies by your daily volume, priced at your model’s input rate, to give a monthly figure. Verbose descriptions are usually the biggest offender, and the calculator makes that cost concrete.

Tips and notes

Trim tool descriptions first — they are free-text and often the largest token sink in a schema while contributing the least to correct tool selection.
Send only the tools relevant to each request rather than the full catalogue; routing the right subset cuts both tokens and tool-selection errors.
Enable prompt caching on the stable tools block so the schema is billed once per cache window instead of on every single call.