LLM Tool-Call Schema Builder
Define a function for LLM tool use once and get a valid schema in the shape each provider expects — Anthropic input_schema, OpenAI tools, or plain JSON Schema. Build it from a form, or infer it from a sample arguments object. Everything runs in your browser.
New to this? Read the tool schema builder guide →
Output is validated as you type. Everything runs locally in your browser — nothing is uploaded. Confirm field names against your provider's current tool-use docs.
How to use the schema builder
- Name your function and describe what it does — the model reads the description to decide when and how to call it, so make it specific.
- Add each parameter with a type, a description and whether it is required. Add an enum for string parameters that accept a fixed set of values.
- Switch the output between Anthropic, OpenAI and plain JSON Schema, then copy it into your API request. Or paste a sample arguments object and let the tool infer a schema for you.
Why the schema matters
When you give a model tools, the schema is the contract: it tells the model which functions exist, what arguments they take and which are required, and it constrains the JSON the model must produce. Good descriptions and tight types (enums, required fields) dramatically improve how reliably the model calls your function with valid arguments. A vague schema leads to missing fields and malformed calls.
Anthropic vs OpenAI shape
Both providers use JSON Schema for the parameters; only the wrapper differs. Anthropic expects { name, description, input_schema }; OpenAI wraps it as { type: "function", function: { name, description, parameters } }. This tool emits the same underlying schema in either shape, so you can target both from one definition. Field names occasionally change, so check the current tool-use documentation for your provider.
Frequently asked questions
What is an LLM tool-call schema?
When you give a model tools (function calling), you send a schema describing each function — its name, a description, and a JSON Schema for its arguments. The model uses it to decide when to call the function and to produce a structured arguments object that matches. Good descriptions and tight types make those calls far more reliable.
What is the difference between the OpenAI and Anthropic formats?
Both describe arguments with the same JSON Schema; only the wrapper differs. Anthropic expects { name, description, input_schema }, while OpenAI wraps it as { type: "function", function: { name, description, parameters } }. This tool emits either from one definition.
Can it generate a schema from example data?
Yes. Paste a sample arguments object in Infer mode and it derives a JSON Schema — string, integer, number, boolean, array and nested object types — with every field marked required as a starting point.
Is anything uploaded?
No. Schema generation is pure JSON processing in your browser. Provider field names change occasionally, so check the current tool-use docs before shipping.