Function-Calling Validator

Validate tool-call JSON against your schema.

How to use Function-Calling Validator

  1. Paste the JSON Schema. Put the schema for the tool's arguments in the JSON Schema box. A sample schema is pre-filled.

  2. Paste the tool call. Put the call in the Tool call JSON box. It needs a name and arguments, either at the top level or inside a function object.

  3. Read the result. You see either a valid message, a malformed tool call error, or a list of schema violations with the path of each failing value.

  4. Fix and re-check. Edit either box. The validation re-runs as you type.

Limits and privacy

  • Supported schema keywords: type (including integer and type arrays), properties, required, additionalProperties, items, enum, const, minimum, maximum, minLength, maxLength, pattern, minItems and maxItems.
  • Not supported: $ref, oneOf, anyOf, allOf and conditional schemas. Those keywords are ignored rather than reported.
  • The arguments field can be a JSON string, as OpenAI-style calls use, or an already-parsed object.
  • The envelope check requires a non-empty name and an arguments value, and reports an arguments string that is not valid JSON.
  • Validation runs in your browser. Nothing is uploaded.

Frequently asked questions

How do I check that a model's tool call matches my schema?

Paste your parameters schema on the left and the tool call on the right. The tool parses the arguments and lists every place they break the schema.

Which tool call formats does it understand?

An object with name and arguments, or one with a function object containing name and arguments. Arguments can be a JSON string or an object.

Does it support the full JSON Schema spec?

No. It covers the common subset used in function-calling schemas and ignores keywords such as $ref, oneOf, anyOf and allOf.

Does it call any model or API?

No. It only validates the JSON you paste, in your browser.

What's next