Skip to content

AI Tool Schema Builder & Validator

Build JSON Schema and function definitions for OpenAI, Anthropic or Gemini.

Found a problem, or something that could be better? We read every message and we fix things quickly.

AI tool schema builder creates tool definitions from sample JSON or an editable JSON Schema subset.

Export OpenAI Responses, Anthropic or Gemini declarations and validate supplied argument JSON locally. No model runs and no API acceptance is claimed.

How to use

  1. Enter a function name, description and sample JSON or supported schema.
  2. Choose the provider and optionally paste argument JSON to check.
  3. Generate, inspect inferred fields and test the definition with your provider.

Frequently asked questions

Which schema keywords work?

The supported subset is type, description, properties, required, additionalProperties, items, enum and anyOf. Other keywords are rejected.

How does OpenAI strict mode change optional fields?

Every object disallows extra properties and requires all declared fields. Previously optional fields become nullable; review this semantic change.

Can a sample describe every valid input?

No. Inference sees only supplied values. Empty arrays need an explicit item schema, and omitted fields cannot be inferred.

Does valid mean the model will call it?

No. AJV checks supplied argument data locally. Provider limits, model behavior, permissions and actual tool execution need separate tests.

Provider wrappers

OpenAI exports the Responses API function shape, Anthropic uses input_schema and Gemini uses parametersJsonSchema. These are definitions to integrate, not complete API requests.

Bounded validation

Use at most 100000 characters per JSON input, 500 schema nodes and 12 nesting levels. Remote references, schema code and model calls are not supported.