JSON Schema Generator (from JSON samples)

A JSON Schema describes what valid JSON looks like — which keys exist, their types, which are required — so APIs, config files and forms can be validated automatically. Writing one by hand for a large response is tedious. Paste one or more samples and get a schema for draft 2020-12, 2019-09, 07, 06 or 04: nested objects and arrays are described recursively, keys present in every sample become required, whole numbers become integer, and common string formats such as date-time, email, uri and uuid are detected.

How to use this tool

1. Paste JSON: an array of samples, a single document, or one document per line.
2. Tell the tool how the samples are arranged.
3. Choose the draft your validator uses.
4. Optionally add a title and $id, and choose required keys, additionalProperties: false, format detection and enums.
5. Copy the schema and refine it — add descriptions, ranges and patterns that samples cannot show.

Formula

null → "null" · true/false → "boolean" · 3 → "integer" · 2.5 → "number" · "…" → "string" (+ format) · [ … ] → "array" + items · { … } → "object" + properties

Values seen at the same place are merged: integer and number become number, a value that is sometimes null gets a type list such as ["string", "null"], and objects or arrays mixed with other kinds become anyOf. A key is required only if every object at that place has it. A format is added only if every string at that place matches it and the draft defines it (date and time from draft-07, uuid and duration from 2019-09).

Example

The three sample users give an object schema with 7 properties: id (integer), email (string, format email), role (string), joined (string, format date-time — "+09:00" offsets are fine), tags (array of strings; one sample has an empty array), manager (null, seen once) and score (number, seen once). id, email, role, joined and tags are required because every sample has them; manager and score are not.