JSON Viewer and Validator

Open a JSON file and browse it as a collapsible tree, with validation, search and structural statistics. The file is read by your browser; it is not uploaded, and no network request is ever made because of anything inside it.

Loading the JSON Viewer…

What this tool does

This viewer accepts any single valid JSON value — an object, an array, a string, a number, a boolean, or null — and shows it as a tree you can expand and collapse. Unlike this site's CSV and Parquet tools, it never touches the local data engine: the file is parsed directly by your browser's own JSON parser, so there is no engine to start and no extra download before you can use it.

One document, not a stream of records

This tool shows exactly one JSON document. If your file is newline-delimited JSON (JSON Lines — multiple independent JSON values, one per line, often named .jsonl or .ndjson), it is detected and rejected with a clear message rather than partially parsed or silently treated as one giant array: use JSON Flattener or JSON to CSV instead, which are built for row-oriented data like that.

Statistics and search

The statistics panel reports structural facts computed directly from your document: character count, maximum nesting depth, how many objects/arrays/strings/numbers/booleans/nulls it contains, and, when the top level is an array of objects, how many of them have each key — a quick, honest way to see whether the records share the same shape without a full profiling pass. Search matches any key or value, case-insensitively, and expands the branches needed to show every match.

A real limitation: number precision

Numbers are parsed as standard JavaScript numbers. This means an integer larger than 2⁵³ can lose precision, a number too large to represent at all (like 1e400) becomes invalid when reformatted, -0 becomes 0, and formatting quirks like a trailing .0 or scientific notation are not preserved exactly. This is disclosed here rather than hidden: if exact preservation of unusual numeric literals matters, this is not the right tool for that file.

Untrusted content stays inert

Every key and value is rendered as plain text, never as HTML, so a string containing markup or a script tag is shown exactly as written and never executed. A document nested far beyond any reasonable real-world structure is refused outright rather than risking a crashed tab.

Large files

Because this tool holds the whole parsed document in your browser's memory (there is no local database engine involved, unlike the CSV and Parquet tools), its size limits are lower: expect a warning above roughly 25 MB, a stronger warning above roughly 75 MB, and a refusal above roughly 150 MB. These numbers come from measuring real files, not a guess.