JSON Schema generator — draft-07 from a sample

What a generated schema gives you

A JSON Schema is a machine-readable description of what a valid document looks like. Unlike TypeScript types, it survives to runtime, so a validator can reject a malformed payload at the boundary of your system rather than letting it propagate.

Writing one by hand for a large document is slow. Generating from a real sample gets the structure right immediately, and leaves you to add the parts that require judgement: constraints, formats, and which fields are genuinely required rather than merely present in your sample.

Worked example: a user record

A sample with mixed numeric types, an array, and a nested object produces:

1. Paste a representative document into the left pane.
2. Read the draft-07 schema on the right, with $schema already declared.
3. Check the required array. Everything present in your sample is listed — remove any field that is genuinely optional in your API.
4. Add constraints by hand: minLength on strings, minimum on numbers, format on emails and dates, enum where a field has a fixed set of values.

What to tighten by hand

• Formats. A string holding an email is just "string" to the generator. Adding "format": "email" or "format": "date-time" gets you real validation.
• Enums. A status field that is always one of three values should be an enum, not an open string. The generator cannot know the full set from one sample.
• Ranges. minimum, maximum, minLength and maxLength encode business rules that no sample can reveal.
• additionalProperties. Draft-07 allows unknown keys by default. Setting it to false makes the schema strict, which is usually what you want for an API contract.
• Required. The generator marks everything in your sample as required. That is almost always too strict for a real API.

When not to use this

Related workflow

For compile-time safety in a TypeScript codebase, the JSON to TypeScript converter generates interfaces from the same sample. The two are complementary: types catch mistakes while you write code, schemas catch bad data while it runs.

To build a schema from many records rather than one, convert your log or dataset with the JSONL converter into an array and paste that — merging across elements produces a far more accurate required list.

The JSON formatter validates a document before you generate from it, and the JSON diff tool helps when comparing two versions of a payload to see which fields actually vary.