vs json.site

json.site and the dozens of similar tools are genuinely convenient. The question is whether convenience is worth sending a payload containing customer records to a server you know nothing about. This page is an honest side-by-side rather than a hit job.

Capability	JSONForge	json.site / typical online tools
Where the JSON is processed	Your browser (Web Worker)	Uploaded to their server
Max file size	100 MB	Usually 1–5 MB
10,000-row array with 1 row inserted	1 added (+ moves shown)	10,000 modified
Reports moved rows	Yes — separate "moved" state	No
Repairs truncated LLM output	Yes, with a per-line change report	No
Inferred JSON Schema	draft-07 and 2020-12	Usually draft-07 only
JSONPath and JMESPath	Both	One or neither
Works offline	Yes (PWA)	No
Sign-up / paywall	None	Often capped or paid
Language	English and 中文	Usually English only

Measured on a 5 MB document in Chrome 130, M1 MacBook Air.

The one difference that matters most

Array alignment. When a row is inserted at the top of a large array, an index-based diff reports every subsequent row as removed and re-added. Matching rows by key reports one addition and N moves — which is the difference between an unreadable result and an actionable one.

Where they are actually better

If you need to share a formatted document with a colleague who will not install anything, a hosted tool with a permalink is simpler than ours. If you work on a locked-down machine that blocks Web Workers, ours will not run. Those are real trade-offs, not footnotes.