JSON Schema Validator

Which drafts are supported?

The validator reads the $schema keyword to pick the draft: draft-04, draft-07 (also used for draft-06), 2019-09 and 2020-12. Without a $schema keyword it assumes 2020-12. The schema itself must be valid JSON.

Reading validation errors

Each problem shows the location of the failing value and the keyword that failed, for example required, type, minimum or additionalProperties, followed by a plain message. When one value breaks several rules you will see one entry per rule, so you can fix them all in one pass.

Start from a generated schema

Generate from JSON infers a schema from your data: property types, required keys (keys present in every object of an array) and array item types. It is a starting point; tighten it by adding keywords such as minLength, enum, pattern or format.

JSON Schema in five minutes

A JSON Schema is a JSON document that describes what other JSON documents must look like. The keyword type says what kind of value is allowed. For objects, properties describes each key, required lists the keys that must be present, and additionalProperties set to false rejects keys you did not list. For arrays, items describes the elements. Numbers can be limited with minimum and maximum, strings with minLength, maxLength, pattern and format, and any value can be limited to a fixed set with enum.

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "required": ["id", "email"],
  "properties": {
    "id": { "type": "integer", "minimum": 1 },
    "email": { "type": "string", "format": "email" },
    "role": { "enum": ["admin", "editor", "viewer"] },
    "tags": { "type": "array", "items": { "type": "string" }, "uniqueItems": true }
  },
  "additionalProperties": false
}

Reading the error list

Every problem is shown with the location of the failing value and the keyword that failed. A required error points at the object that is missing a key, not at the key itself, because the key does not exist. A type error points at the value that has the wrong type; the most common cause is a number sent as a string. An additionalProperties error means the object contains a key your schema does not allow, which is often a typo in the key name.

When one value breaks several rules you see one entry per rule. For example, the value -3 for an age with a minimum of 0 and a type of integer fails only minimum, but the string "3" fails type and also cannot be compared for minimum, so fix the type first.

Drafts and what changed between them

The validator reads the $schema keyword to choose a draft. Draft-04 uses exclusiveMinimum and exclusiveMaximum as booleans next to minimum and maximum; draft-06 and later made them numbers. Draft-07 added if, then and else. Draft 2019-09 introduced $defs for reusable parts (older drafts used definitions) and unevaluatedProperties. Draft 2020-12 changed how tuples are written: prefixItems replaces the array form of items.

If your schema was written for one draft and behaves oddly, set $schema explicitly. Without it, 2020-12 is assumed.

Combining and reusing schemas

allOf requires a value to match every sub-schema, anyOf at least one, and oneOf exactly one. oneOf is the cause of many confusing errors, because a value that matches two branches fails it. Use $ref with $defs to reuse a definition, for example "$ref": "#/$defs/address". References to other files or URLs are not fetched by this tool, because it never makes network requests with your data; paste the referenced definitions into one schema instead.

format is a hint in the specification and validators may treat it as an annotation only. This tool checks common formats such as email, date-time, uri and uuid, but it is worth testing the formats you rely on with your production validator as well.

Using the schema elsewhere

Once a schema works here you can use it in your own code or pipeline with a validator library for your language, and in editors that understand JSON Schema to get completion and inline errors for configuration files. Start from the Generate from JSON button, which writes a first schema from your data, then tighten it by hand with the keywords above.