What instance types does JSON Schema define?

JSON Schema validates against the JSON data model: string, number, integer, boolean, object, array and null. The type keyword restricts an instance to one or more of these. Type-specific keywords such as minItems or properties only take effect when the instance is the matching type.

What is the difference between draft-07 and 2020-12?

2020-12 is a newer draft that splits array validation into prefixItems (tuples) and items (the rest), adds $defs as the standard definitions container and refines reference resolution. Draft-07 introduced if/then/else. Many tools support both, but keyword availability differs, so target one draft explicitly.

Does the type keyword accept multiple values?

Yes. type can be a single string like "string" or an array such as ["string", "null"], which is the common way to allow a nullable value. The instance passes if it matches any listed type.

Is format actually enforced?

By default format is an annotation, not an assertion, so an invalid email may still validate unless your validator has format assertion enabled. For guaranteed checks combine format with a pattern or enable strict format validation in your tooling.

Does this tool send my schema anywhere?

No. The reference is a static table filtered in your browser. Nothing you type is uploaded, logged or stored.

JSON Schema Type Reference

Email me this result

Get this tool's output sent to your inbox, plus one useful tool a week. No spam, unsubscribe any time.

Every JSON Schema keyword, grouped by what it constrains

JSON Schema describes the shape of JSON data using a vocabulary of keywords. This reference lists the instance types and the validation keywords that apply to each — string, numeric, object, array and the boolean combinators — for the draft-07 and 2020-12 drafts, with a short example for every entry.

How it works

A JSON Schema is itself a JSON object. The type keyword pins the instance to one of the seven JSON types, and type-specific keywords then add constraints. Keywords that do not match the instance type are simply ignored, which is why a schema can combine, say, string and array constraints safely.

{
  "type": "object",
  "required": ["id", "tags"],
  "properties": {
    "id": { "type": "integer", "minimum": 1 },
    "tags": { "type": "array", "items": { "type": "string" }, "uniqueItems": true }
  },
  "additionalProperties": false
}

Boolean combinators — allOf, anyOf, oneOf and not — let you compose subschemas, and if/then/else applies them conditionally.

Tips and notes

Prefer additionalProperties: false to catch typos in property names during validation.
In 2020-12 use prefixItems for tuples and items for the remaining elements; in draft-07 a items array did the positional job.
$ref plus $defs keeps large schemas DRY by referencing reusable subschemas.
format is an annotation unless your validator enables format assertion — pair it with pattern when you need a hard guarantee.