JSON Schema Batch Validator
Check a whole array or NDJSON stream of records against one JSON Schema. See a pass/fail summary and every failing record's errors with JSON Pointer paths — all locally, no upload.
About this tool
Paste one JSON Schema and a whole batch of records, and this tool validates every record
against that schema in one pass. The batch can be a JSON array of records, a single JSON value
(treated as one record), or an NDJSON stream (one JSON value per line). You get a PASS/FAIL
verdict, a summary of how many records passed and failed, and — for each failing record — the exact
errors, each with a JSON Pointer path, the schema keyword that rejected the value, and a
human-readable message.
It runs entirely in your browser as WebAssembly: nothing is uploaded, and no network $refs are
fetched. Choose the JSON Schema draft label explicitly, or leave it on Auto to detect the draft
from the schema's $schema keyword (falling back to Draft 2020-12). Validation is intentionally
local and covers the common draft-agnostic keyword subset listed in the limits below.
Worked example
Schema
{"$schema":"http://json-schema.org/draft-07/schema#",
"type":"object","required":["id","name"],
"properties":{"id":{"type":"integer"},"name":{"type":"string"}}}
Records (JSON array)
[{"id":1,"name":"Ada"},{"id":"2","name":"Grace"},{"name":"Kay"}]
Result
Batch validation: FAIL
Draft: draft7 (from $schema) Input: json Records: 3 (1 passed, 2 failed)
Total errors: 2
Record #1 — 1 error
/id: "2" is not of type "integer"
Record #2 — 1 error
(root): "id" is a required property
Record 0 passes. Record 1 fails because id is a string, not an integer. Record 2 fails because
the required id property is missing. Switch Output format to JSON for the same result as a
machine-readable object you can pipe into another tool.
FAQ
What batch formats can I paste into the records box?
Three shapes are accepted. A JSON array [{...},{...}] validates each element as a record. A
single JSON value (object, number, string, …) is validated as one record. NDJSON (also
called JSON Lines) is one JSON value per line, with blank lines skipped. Leave Records format on
Auto to try whole-document JSON first and fall back to NDJSON, or force json / ndjson when
you know the shape.
Which JSON Schema drafts are supported, and how does auto-detect work?
Draft 4, Draft 6, Draft 7, Draft 2019-09, and Draft 2020-12 can be selected for reporting. On
Auto, the tool reads the schema's $schema URI and picks the matching draft label; if there is
no $schema (or it isn't recognized) it falls back to Draft 2020-12. The validator applies the
same local keyword subset for each draft, so use a full validator when you need draft-specific
features such as oneOf, conditionals, unevaluated properties, or format assertions.
What does each reported error contain?
Every error carries three things: a JSON Pointer path to the failing value (`` empty / shown as
(root) for the record itself, /age for a member, /tags/0 for the first array item), the
schema keyword that rejected it (type, required, minimum, enum, pattern, …), and a
message. Errors within a record are sorted by path so the output is stable. The per-record and
total error counts are always exact, even when the listing is capped.
What is "Max errors to list" for?
It caps how many individual errors are printed across the whole batch (default 50) so a huge, badly
broken batch stays readable. Only the listing is capped — the per-record error_count and the
batch total_errors are still counted in full, and the report is marked truncated when more
errors exist than were shown.
Does it fetch remote schemas or `$ref` URLs?
No. The tool is fully local: it never makes network requests, so external HTTP $refs are not
resolved. Internal $refs inside the schema you paste (references within the same document) are
resolved normally. If your schema depends on remote definitions, inline them before validating.
Limits & edge cases
- Local only — no network, so remote/HTTP
$refs are not fetched (internal$refs work). - Keyword subset — validates
type,required,properties,additionalProperties,enum,const, numeric bounds, string length,pattern,items, and internal$ref. Complex composition/conditional keywords andformatassertions are treated as annotations. - An empty batch (
[]) is reported as valid with 0 records. - Passing records are counted in the summary but not listed individually; only failing records get a detailed error breakdown.
formatkeyword handling follows the selected draft's rules; when in doubt, pick the draft explicitly rather than relying on auto-detect.- Very large batches are still validated in full, but the error listing is bounded by Max errors to list.
Developer & Automation Access
Run it from the terminal
Same engine as this page, headless — via the gizza CLI:
gizza tool json-schema-batch-validate '{"type":"object","required":["id","name"],"properties":{"id":{"type":"integer"},"name":{"type":"string"}}}' 'records=[{"id":1,"name":"Ada"},{"id":"2","name":"Grace"},{"name":"Kay"}]'New to the CLI? Get gizza →
Open it by URL
Pre-fill and auto-run this tool with query parameters — the names match the API/CLI:
https://gizza.ai/tools/json-schema-batch-validate/?schema=%7B%22type%22%3A%22object%22%2C%22required%22%3A%5B%22id%22%2C%22name%22%5D%2C%22properties%22%3A%7B%22id%22%3A%7B%22type%22%3A%22integer%22%7D%2C%22name%22%3A%7B%22type%22%3A%22string%22%7D%7D%7D&records=%5B%7B%22id%22%3A1%2C%22name%22%3A%22Ada%22%7D%2C%7B%22id%22%3A%222%22%2C%22name%22%3A%22Grace%22%7D%2C%7B%22name%22%3A%22Kay%22%7D%5D&input_format=auto&draft=auto&max_errors=50&output=textMachine-readable descriptor: tool.json — title + parameters JSON Schema for agents.