Free · no signup · browser-local

JSON Schema 2020-12 Generator from Example JSON

Infer a JSON Schema from nested sample data, choose required and additional-property rules, then validate another JSON document locally with Ajv.

Browse all tools →

Local workspace · no model calls · nothing saved

Make it yours

Schema title, description and identifier · optional

titledescription$idAbsolute HTTP(S) or URN, without fragments or credentials. An identifier only; no address is requested.

Sample JSON

{
  "person": {"name":"Ada","active":true},
  "tags": ["research","writing"],
  "score": null
}

Observed object keys are requiredAllow additional object properties

Test data · optional, validated locally

{
  "person": {"name":"Ben","active":false},
  "tags": ["planning"],
  "score": null
}

Test JSON matches the schema

Detailed validation report

{
  "dialect": "2020-12",
  "scope": "Structural validation against a sample-derived schema; business rules and string formats are not inferred or validated.",
  "valid": true,
  "errorLimit": 50,
  "errorCount": 0,
  "errorsTruncated": false,
  "errors": []
}

Draft 2020-12. Empty arrays accept any item; mixed sample shapes use anyOf. No business rules or format assumptions are inferred. Limit: 100 KB input, depth 20, 5,000 nodes; generated schema up to 250 KB.

Your output

schema.json

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "person": {
      "type": "object",
      "properties": {
        "name": {
          "type": "string"
        },
        "active": {
          "type": "boolean"
        }
      },
      "required": [
        "name",
        "active"
      ],
      "additionalProperties": false
    },
    "tags": {
      "type": "array",
      "items": {
        "type": "string"
      }
    },
    "score": {
      "type": "null"
    }
  },
  "required": [
    "person",
    "tags",
    "score"
  ],
  "additionalProperties": false
}

Getting started

Paste sample JSON, choose required keys and additional properties, then test a separate JSON document.

Before you use the result

A single sample cannot infer business rules, enums or all valid shapes. Empty arrays accept any items; observed array shapes become alternatives.

Try an example

Infer an object with a nested person and mixed array, then test a missing key or incorrect type.

How it works

Recursive type inference creates an actual schema; Ajv validates test data locally. Limits are 100 KB, 20 nesting levels and 5,000 nodes.

Frequently asked questions

Infer a draft 2020-12 JSON Schema from sample JSON and validate another value locally.

How do I use JSON Schema Builder?