JSON Schema

JSON Schema is a declarative language — written in JSON — for describing what a valid JSON document looks like. A schema lists required keys, value types, numeric ranges, string patterns, enumerated values, and how parts of a document can compose with allOf, anyOf, oneOf, and $ref. Given a schema and a candidate document, a JSON Schema validator produces either “valid” or a list of specific violations.

A minimal example

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "required": ["name", "email"],
  "properties": {
    "name":  { "type": "string", "minLength": 1 },
    "email": { "type": "string", "format": "email" },
    "age":   { "type": "integer", "minimum": 0 }
  },
  "additionalProperties": false
}

Given this schema, {"name": "Ada", "email": "ada@example.com"} is valid. {"name": ""} fails (missing email, and name is empty). {"name": "Ada", "email": "ada", "role": "admin"} fails twice (email format, and the extra role field).

Drafts

The specification has evolved through several drafts. The ones in wide use:

• Draft-04 — old but still encountered in legacy systems.
• Draft-07 — widely deployed.
• 2019-09 and 2020-12 — the current set, including $defs and improved items / prefixItems semantics.

Always include $schema to tell validators which rules apply.

Use cases

• Rejecting malformed API requests at the boundary, before they reach business logic.
• Generating API documentation automatically (OpenAPI embeds JSON Schema).
• Powering form UIs (libraries like react-jsonschema-form render a schema directly into a form).
• Generating type definitions for TypeScript, Rust, Go, and others from a single source of truth.
• IDE autocomplete and validation for config files (VS Code reads $schema from a JSON file and highlights errors in real time).

For a longer tour, see JSON Schema Basics.