JSON Schema Generator

Updated October 18, 2026Free mock REST & WebSocket APIs

Paste sample JSON to generate a JSON Schema instantly — types, nested objects, arrays, required fields and string formats inferred for you. One click turns it into an OpenAI structured-outputs response_format or a Claude tool input_schema.

Sample JSON

Paste, type, or drop a file onto the box.

JSON Schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "title": "Invoice",
  "type": "object",
  "properties": {
    "invoice_id": {
      "type": "string"
    },
    "issued": {
      "type": "string",
      "format": "date"
    },
    "customer": {
      "type": "object",
      "properties": {
        "name": {
          "type": "string"
        },
        "email": {
          "type": "string",
          "format": "email"
        }
      },
      "required": [
        "name",
        "email"
      ]
    },
    "lines": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "sku": {
            "type": "string"
          },
          "description": {
            "type": "string"
          },
          "qty": {
            "type": "integer"
          },
          "unit_price": {
            "type": "number"
          },
          "discount": {
            "type": "number"
          }
        },
        "required": [
          "sku",
          "description",
          "qty",
          "unit_price"
        ]
      }
    },
    "paid": {
      "type": "boolean"
    },
    "notes": {
      "type": "null"
    }
  },
  "required": [
    "invoice_id",
    "issued",
    "customer",
    "lines",
    "paid",
    "notes"
  ]
}

Title

Detect formats (date-time, email, uri…)additionalProperties: falseTop-level array = several samples of one object

🔒 Generated in your browser — your JSON is never uploaded.

How to generate a JSON Schema from JSON

Paste a sample JSON document (or an array of several samples) or open a .json file.
Choose the output: plain JSON Schema (draft 2020-12 or draft-07), OpenAI structured output, or Claude tool definition.
Adjust required fields, format detection and additionalProperties, then copy or download the result.

How the schema is inferred

Sample value	Schema
`42`	`{"type": "integer"}`
`4.5` (or 42 and 4.5 in different items)	`{"type": "number"}`
`"2026-10-18T09:30:00Z"`	`{"type": "string", "format": "date-time"}`
`"ada@example.com"`	`{"type": "string", "format": "email"}`
`null`	`{"type": "null"}`
Array of objects	Items merged; a key is `required` only if every item has it

JSON Schema for LLM structured outputs

OpenAI's Structured Outputs (strict: true) accept a subset of JSON Schema: every property must be listed in required, objects need additionalProperties: false, and optional fields are expressed as a union with null. The “OpenAI structured output” option applies these rules for you:

{
  "type": "json_schema",
  "json_schema": {
    "name": "Invoice",
    "strict": true,
    "schema": {
      "type": "object",
      "properties": {
        "invoice_id": { "type": "string" },
        "discount":   { "type": ["number", "null"] }
      },
      "required": ["invoice_id", "discount"],
      "additionalProperties": false
    }
  }
}

For Claude, put the schema in a tool's input_schema (or use structured outputs with output_config.format); add strict: true on the tool to guarantee schema-valid arguments.

Tips

Give several samples (an array, with the “several samples” option) so optional fields and nullable values are detected correctly.
Add description fields by hand after generating — models use them as instructions.
Need TypeScript or Zod instead? Use JSON to TypeScript or JSON to Zod.

Related tools

Frequently asked questions

Is my JSON uploaded?

No. The schema is generated in your browser; your JSON never leaves your device.

Which JSON Schema draft should I use?

Draft 2020-12 is current and supported by Ajv, OpenAPI 3.1 and most validators. Choose draft-07 for older tools that don't understand 2020-12.

Why does the OpenAI version make fields nullable?

Strict structured outputs require every property to be required. A field that's missing in some samples is therefore kept required but allowed to be null, which is how OpenAI recommends expressing optional values.

Can it detect enums?

Not automatically — a sample can't tell an enum from free text. Replace a string type with "enum": ["paid", "refunded"] when the values are fixed.

Your business here? Sponsor this tool from $16/week. See placements