JSON Schema Validator - Draft 2020-12 Contracts in the Browser

Schema - integer age, minimum 0, required

{
  "type": "object",
  "properties": {
    "age": { "type": "integer", "minimum": 0 }
  },
  "required": ["age"]
}

Instance - valid (default)

{"age": 21}

Result: Valid (2020-12). 21 is an integer and is greater than or equal to 0. The required property is present.

Instance - type failure

{"age": "21"}

Result: Invalid. Ajv reports something like /age must be integer. JSON Schema does not coerce strings to numbers. That is the whole point of a contract: "21" and 21 are different values.

Instance - minimum failure

{"age": -1}

Result: Invalid. The type is still integer, but minimum rejects negative ages. A syntax validator would call this document perfectly legal JSON. Only a schema can say “legal JSON, illegal age.”

Checking API request bodies before you send them

OpenAPI and many internal style guides publish a JSON Schema for POST and PUT bodies. Pasting a payload here catches missing required keys, wrong types, and out-of-range numbers before the request leaves your machine. That is faster than reading a 400 from a staging cluster, and it does not require a mock server.

Catching string-vs-number bugs in form JSON

HTML forms and some client libraries serialize numbers as strings. A syntax check will happily accept {"age":"21"}. A schema with type: integer will not. This is one of the most common production mismatches between a TypeScript interface that says number and a JSON body that says string.

Governing configuration files

Tools that load JSON config - linters, deploy manifests, feature-flag dumps - often ship a schema even when they also parse with JSON.parse. Validating a local config against that schema finds unknown enums, missing required blocks, and accidental extra keys (once you set additionalProperties false) without starting the whole application.

Teaching JSON Schema without installing Ajv

Draft 2020-12 vocabulary is easier to learn when you can change one keyword and see the error path update immediately. Students can keep the default age schema, then add maximum, enum, or nested properties without a Node toolchain. The dialect label in the result panel shows whether compile stayed on 2020-12 or fell back to draft-07.

Reviewing vendor schemas before you adopt them

Third-party APIs sometimes publish a schema that is draft-07 while your stack assumes 2020-12, or the reverse. Compiling here is a cheap compatibility check. If 2020-12 throws and draft-07 succeeds, you know you need an older validator in CI or a schema upgrade - not a mystery production failure.

Closing objects that should not grow extra keys

The default sample is an open object: extra properties pass. Many APIs want the opposite. Adding additionalProperties: false turns a typo such as agge into a validation error instead of a silently ignored field. That is a schema job. A formatter cannot see the typo; a syntax validator cannot see it either.

Linting format keywords locally

ajv-formats assists email, uri, and date-time. That is useful when you want a second pair of eyes on a payload that already parses. It is not a replacement for a dedicated email RFC library. Treat a format failure as a hint, then confirm against the spec your product actually promised.

Separating “is JSON” from “is our JSON”

Incident write-ups often confuse a parse error with a schema error. If the document will not parse, this page never reaches Ajv. If it parses and still fails, the report is a contract failure. Keeping those two tools separate - this page versus the JSON validator - makes the postmortem accurate.

What is JSON Schema validation?

Syntax checks ask whether a document is JSON. Schema checks ask whether the parsed value matches a contract: types, required keys, ranges, enums, nested objects, and optional format hints. This page compiles with Ajv using draft 2020-12 first, then retries draft-07 if compile throws. It is not a pretty-printer and not a syntax-only parser.

What does the default sample do?

The default instance is {"age":21} and the default schema is an object that requires age as an integer with minimum 0. Ajv reports valid. Change age to -1 and the minimum keyword fails. Change age to the string "21" and the type keyword fails. That is the worked check for this tool.

Is this the same as a JSON validator?

No. A JSON validator only runs a parse and reports the first line and column if RFC 8259 fails. This page assumes both the instance and the schema are JSON, then applies the schema. If either side will not parse, fix that on the JSON validator first, then come back here.

Which JSON Schema draft is supported?

Primary compile is JSON Schema draft 2020-12. If that fails because the schema is an older draft-07 document, the tool retries with Ajv’s draft-07 validator. Draft-04 exclusiveMinimum-as-boolean quirks are not fully emulated. Remote $ref HTTP loads are not fetched.

Are format assertions such as email enforced?

ajv-formats is loaded, so keywords like email, uri, and date-time are assisted. Treat format as a useful lint, not a complete RFC implementation of every format string. Some ecosystems treat format as an annotation rather than an assertion.

Does schema validation upload my data?

No. Ajv runs in the browser. The instance and schema stay in local storage for up to 30 days so you can resume a draft. Nothing is posted to a server. Do not paste production secrets if other people share the device.

What happens if the instance or schema is not valid JSON?

The schema engine never runs. You get a parse error that names the side that failed (Instance or Schema) plus line and column from parseJsonDocument. Repair small broken drafts on JSON repair, or check syntax on the JSON validator, then paste the clean documents here.

Does additionalProperties reject extra keys by default?

No. The default sample schema does not set additionalProperties to false, so extra keys are allowed. Add that keyword when you want a closed object. That difference is a common source of “it validated but the API still rejected it” surprises.

Can I validate against a remote $ref?

Not over the network. Fetching a remote schema would be a side effect this page does not perform. Inline the referenced definitions, or resolve $ref locally before pasting. The compile error will name the problem if a $ref cannot be resolved in-document.

Is this JSON Schema validator free?

Yes. No signup, no usage cap, and no premium dialect. Validate as many instance and schema pairs as you need. Processing stays on your device.