Skip to the tool
GigaJSON

JSON Schema Generator

All tools
Format and check
View and query
Convert
JSON to code
Encode and decode
Sample files
Guides Open GigaJSON
Input JSON
Output JSON Schema

The JSON Schema output appears here as you type.

Ready Runs in your browser

Paste example JSON and get a JSON Schema that describes it: the type of each field, which keys are required, which can be null, and the shape of nested objects and lists. Give it several records in a list for the best result, since a schema made from one record marks every key as required. To check data against the schema, use the JSON Schema Validator. For TypeScript interfaces instead, use JSON to TypeScript.

How to use it

  1. Paste JSON (several examples in a list work best) or open a file.
  2. The schema appears as you type.
  3. Copy it or download a .schema.json file, or open the data in GigaJSON to validate it against the schema.

Read more: A JSONLint alternative for large JSON files · How to create a JSON Schema from JSON

Private, and made for big files

This tool runs in your browser, in a background worker: nothing you paste or open is uploaded, and nothing is kept after you leave. JSON and XML input can be up to 1 GB, depending on available memory. Processing runs in a background worker; large analyses can take a minute or more. Already loaded features keep working if your connection drops. Loading another tool or jq for the first time needs a connection. When one tool isn't enough, GigaJSON opens your data with a tree, a table, JSONPath/JMESPath/jq queries, transforms, diff and history.

How JSON becomes a JSON Schema

Two users and a map of teams:

{
  "users": [
    { "id": 1, "name": "Ada", "score": 9.5, "tags": ["admin"], "manager": null },
    { "id": 2, "name": "Linus", "score": 7, "tags": [], "manager": 1, "email": "linus@example.com" }
  ],
  "teams": { "101": { "lead": "Ada" }, "102": { "lead": "Linus" } }
}

give this schema (indentation shortened here):

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "users": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "id": { "type": "integer" },
          "name": { "type": "string" },
          "score": { "type": "number" },
          "tags": { "type": "array", "items": { "type": "string" } },
          "manager": { "type": ["null", "integer"] },
          "email": { "type": "string" }
        },
        "required": ["id", "name", "score", "tags", "manager"]
      }
    },
    "teams": {
      "type": "object",
      "additionalProperties": {
        "type": "object",
        "properties": { "lead": { "type": "string" } },
        "required": ["lead"]
      }
    }
  },
  "required": ["users", "teams"]
}
In the dataIn the schema
id is 1 and 2"integer". A field with both 9.5 and 7, like score, is "number".
manager is null once and 1 once["null", "integer"]
email is in one user onlyListed in properties, left out of required
Every item of users is an objectOne items schema that merges all of them
teams has number-like keysadditionalProperties, so new team IDs stay valid
A list mixing typesanyOf with one entry per type

Objects keyed by IDs count as maps when the keys are numbers, or when there are at least eight keys whose values all have the same fields. Anything else is read as a fixed set of properties.

Questions

How are required keys decided?

A key that appears in every object of a list is required. One that's missing from some of them is optional.

Which JSON Schema version is it?

It declares draft 2020-12, and sticks to keywords that earlier drafts share (type, properties, items, required).

Does it detect formats like email or date-time?

No. A string is typed as string, and no enum, pattern, minimum or format is guessed from a few examples. Add the ones you want by hand, such as "format": "email".

Can I validate JSON against the schema?

Yes. Paste both into the JSON Schema Validator, or open the data in GigaJSON and use Types, then Validate. Both check against draft 4, 7, 2019-09 or 2020-12 and list each error with its path.

Why is a field typed as ["null", "integer"]?

The field was null in at least one example and a whole number in another. If null should not be allowed, remove "null" from the list.

Is my data uploaded?

No. The tool runs in your browser, in a background worker. Nothing you paste or open goes to a server, and nothing is kept after you leave the page.

All tools