Webhook HMAC Signer
Sign and verify GitHub, Stripe, and Slack webhook signatures locally.
OpenStart a schema from real data instead of a blank file.
Use a complete, valid example document.
Types, properties, items and required keys are filled in.
Trim required lists and widen types that can vary.
Copy the schema or download the file.
Your team wants to validate incoming config files in CI, and nobody has written a schema yet. The quickest start is to take a known-good file, generate a schema from it, and then loosen the parts that are allowed to vary.
Paste a JSON document and the generator returns a schema that declares the 2020-12 draft in $schema and carries the title GeneratedSchema. Objects get a properties map and a required list containing every key that appeared. Whole numbers are typed integer and others number, strings, booleans and null get their own types, and arrays describe their items using the first element. An empty array produces an items schema of {}, which accepts anything. Output updates as you type, and Download saves it as schema.json.
A schema inferred from one example is strict in some places and loose in others. Every key is required, even ones your system treats as optional, and an array whose first item differs from the rest takes the first item's shape. There are no formats, patterns, enums or length limits. Edit the required arrays, add additionalProperties: false if unknown keys should fail, and widen a "null" type to a list such as string and null where a value can be either. Make sure the sample parses with the online JSON checker, then feed the schema to the Zod and TypeBox code generator for runtime checks in TypeScript.
Inference happens in your browser, so the sample can come from production data without being uploaded.
Keyboard shortcut
Answers for this tool. For site-wide help, open the FAQ hub.
Draft 2020-12, declared in the $schema keyword. If your validator expects draft-07, change the $schema URL; the keywords this tool generates mean the same thing in both drafts.
The generator only sees one document, so it assumes every key in it is always present. Remove keys that can be missing from the required arrays.
No. Paste one representative document. Merging the differences between samples, such as fields that only some records have, is a manual edit afterwards.