Infer a JSON Schema (draft 2020-12) from a sample JSON document — types, required keys, and nested structure, all client-side.
Updated
Quick start
Paste a JSON sample, get an inferred schema — nothing leaves your browser.
Drop a representative JSON document into the left pane.
A draft 2020-12 schema with types, properties, and required keys appears on the right.
Copy with the button or ⌘/Ctrl+Enter, then add formats, enums, and ranges by hand.
In-depth guide
Writing a JSON Schema by hand is tedious; inferring a first draft from a real payload is far faster. This tool reads a sample JSON document and produces a draft 2020-12 schema — types, object properties, required keys, and merged array item shapes — entirely in your browser. Use it as a scaffold you then refine with formats, enums, and constraints.
Given a sample value the tool walks its structure and emits a schema fragment for each node: type for scalars, a properties map plus required for objects, and an items schema for arrays. Whole numbers are typed as integer, fractional ones as number. The root is wrapped with the $schema keyword pointing at draft 2020-12.
It saves the structural typing; you add the semantics (formats, ranges, which fields are truly optional).
required.The output is a starting point, not a finished contract. Always review it before validating production data.
required array or supply varied samples.format, enum, and range keywords yourself.{} (anything). Provide at least one element to constrain it.Parsing and inference run with the browser's native JSON parser inside the tab. No document is uploaded, logged, or stored, and there is no eval anywhere in the code path — so even malformed or sensitive JSON stays on your machine. Close the tab and nothing persists.
Use this tool for quick inspection, cleanup, conversion, and copy-paste debugging. Use your IDE formatter, schema tests, or CI checks when the same rule needs to be enforced across a repository.
No. Your JSON is parsed with the browser's built-in JSON.parse and the schema is inferred in memory. Nothing is uploaded, so you can generate schemas from sensitive payloads safely — confirm in DevTools → Network.
Draft 2020-12, the current version, identified by the $schema keyword pointing at json-schema.org/draft/2020-12/schema. The structural keywords used (type, properties, required, items, anyOf) are compatible with most older drafts too.
From the sample values: strings become "string", booleans "boolean", null "null", and numbers become "integer" when whole and "number" otherwise. Objects produce a properties map with every key marked required, and arrays produce an items schema merged across all elements.
The tool merges the schemas of all array elements. Objects are merged by unioning their properties and keeping only keys present in every element as required. When element types genuinely conflict, it emits an anyOf listing each shape, which you can then refine by hand.
Treat it as a strong first draft. Inference cannot know your intent — it will mark every observed key as required and cannot guess formats (email, date-time), enums, or value ranges. Review and tighten the output before using it to validate real traffic.
The tool only sees one sample, so every key in that sample looks mandatory. If a field is optional, remove it from the required array — or paste a sample array containing objects both with and without the field, and the merge step will drop it from required automatically.
Keep exploring
Hand-picked utilities that pair well with the one you're on — all free, client-side, and zero-signup.
Format, validate, and minify JSON with exact line/column parse errors, optional common-fix mode, and one-click copy for API payloads.
Flatten JSON arrays into CSV with custom delimiters.
Validate JSON against a JSON Schema, see errors inline.
Shrink PDF file size without uploading to a server.
Body Mass Index with metric / imperial inputs and WHO category bands.
Combine multiple PDFs into one in your browser.
Split a PDF by pages or page ranges, download as zip.