What is the difference between the OpenAI and Anthropic formats?

OpenAI nests each tool under a "function" object with "parameters", while Anthropic uses a flat tool object with "input_schema". The builder emits the correct shape for whichever provider you pick so you do not have to remember the differences.

Why does the tool description matter so much?

The model reads the description to decide whether and when to call the tool. A vague description leads to missed or wrong calls. The builder reminds you to write descriptions that state purpose and the situations the tool applies to.

How do required parameters work?

Parameters you mark required are listed in the schema's "required" array, telling the model it must supply them. Optional parameters are still defined but absent from that array, so the model may omit them.

Which parameter types are supported?

The standard JSON Schema primitives — string, number, integer, boolean, and array. For anything more complex you can copy the output and extend the schema by hand, since it is valid JSON Schema.

Does this make API calls?

No. It only generates the request payload shape locally in your browser. You take the output and use it in your own backend call, where your API key stays server-side and secure.

Tool-Use Function Calling Prompt Builder

Tool-use function calling prompt builder

Function calling (tool use) is how you let a model trigger real actions — querying a database, hitting an API, doing math reliably. The hard part is getting the schema exactly right, because OpenAI and Anthropic structure their tools differently and a malformed schema fails silently or causes wrong calls. This builder takes a plain-English description of your tool and emits the correctly structured tools array and a starter system prompt for whichever provider you target.

How it works

You enter the tool’s name, a description of when the model should call it, and its parameters — each with a name, type, description, and a required flag. The builder assembles a valid JSON Schema for the inputs and wraps it in the right envelope: OpenAI nests it under function with a parameters schema, while Anthropic uses a flat object with input_schema. It also generates a short system prompt reminding the model when to use the tool. The output is copy-paste-ready JSON you drop into your own API call; nothing is sent anywhere and your API key never touches this page.

Tips and example

Spend your effort on the descriptions — both the tool’s and each parameter’s — because they are the only thing the model reads to decide what to call and with what arguments. “get_weather: returns current weather for a city; call when the user asks about weather or temperature” beats “weather tool” every time. Mark only genuinely required parameters as required, so the model can omit optional context without inventing values. For complex nested inputs, generate the base schema here and extend the JSON by hand — the output is standard JSON Schema, so it composes cleanly with whatever structure your API needs.