Free online JSON tools

99 JSON and JSON Schema utilities · everything runs in this tab · no account required

Paste JSON below and it is formatted and checked as you type. If you need something other than formatting, there are 99 more tools here — converters, a real jq playground, diff and patch, JSON Schema, and type generators for a dozen languages. None of them upload anything: the parsing happens on your machine, in this page.

Indent Sort keys (recursive)

Paste JSON

Formatted

Ready.

This is the formatter in full. Broken input? Repair it. Need it smaller? Minify.

Pick a job

Format / Validate 8 Eight browser-based tools for making JSON readable and making sure it parses: pretty-print, minify, strict RFC 8259 validation, repair, escaping and RFC 8785 canonical form. Convert 23 Convert JSON to and from CSV, YAML, XML, TOML, Excel, NDJSON, MessagePack, CBOR, HTML and Markdown tables. Every converter runs in the browser with no upload. Config formats 8 Convert JSON to and from dotenv, INI, Java .properties and URL query strings, with the nesting and escaping conventions each format actually uses. Transform 18 Reshape JSON without writing a script: flatten and unflatten, sort keys, deep merge, rename or drop keys, dedupe, filter, group and aggregate arrays. Query / Explore 7 Run jq, JSONPath (RFC 9535) and JMESPath against your own JSON in the browser, or explore a large document as a collapsible tree with copyable paths. Diff / Patch 3 Compare two JSON documents semantically, ignoring key order, and generate or apply RFC 6902 JSON Patch and RFC 7396 merge patch operations. JSON Schema 11 Generate a JSON Schema from samples, validate data against it across drafts 04 to 2020-12, diff two schemas for breaking changes, bundle $refs and convert to OpenAPI. Generate types 17 Turn a sample JSON payload into typed code: TypeScript interfaces, Zod schemas, Go structs, Pydantic models, Rust serde types, Java records, Kotlin, Swift, Dart, PHP, SQL DDL and more. Inspect 5 Measure and profile JSON: raw / minified / gzip size with the heaviest keys, shape statistics, int64 precision loss detection, and an offline JWT decoder.

Ten to start with

JSON formatter Indent 2, 4 or tab, optionally sorting keys so two dumps of the same object diff clean. JSON validator Strict RFC 8259, with the line and column of the first thing that broke. JSON repair Trailing commas, single quotes, bare keys, smart quotes — the mess an LLM hands you. JSON → CSV Array of objects to CSV, with a column picker and a choice of how to flatten nesting. JSON → YAML Block or flow style, key order kept, and a warning when a value hits the Norway problem. jq playground Real jq, compiled to WebAssembly — the same filters your shell pipeline runs. Diff Semantic compare: key order and whitespace are ignored, real changes are not. JSON → TypeScript Interfaces or type aliases, with optionality inferred from an array of samples. Schema generator Infer draft 2020-12 from one sample or several — more samples, better optionality. Tree viewer Collapse a large document down to the shape, then copy the path to any node.

Or see all 99 tools on one page — press / anywhere to search them.

What usually goes wrong with JSON

Most visits here start with a parse error, and there are only about six causes worth knowing. Trailing commas and comments are legal in JavaScript and in JSON5, and illegal in JSON — which is why a config file copied out of a tsconfig.json fails. Single quotes and unquoted keys come from pasting a JavaScript object literal. Curly quotes come from a document editor silently replacing ". All four, plus a few more, are what JSON repair exists to undo in one pass.

Two more are subtler because the document parses fine and is still wrong. A 64-bit id such as 9007199254740993 loses its last digits the moment JavaScript reads it, silently, and you find out when a lookup misses — number precision lists every value in a document at risk. And key order is not meaningful in JSON, so two services serialising the same object produce files a text diff calls different; sorting keys first, or using a semantic diff, is what makes the comparison mean something.

Everything runs in your browser

There is no server-side processing on this site. The page loads, and from then on every tool is JavaScript in your tab using your own CPU. That is not a privacy slogan, it is the architecture: there is no upload endpoint to send a document to, which is why you can paste a production payload with real customer records in it, and why the tools keep working after you disconnect.

The limits are your browser's rather than ours. Documents past roughly 50 MB make a textarea sluggish — drop the file onto the page instead of pasting it — and past a few hundred megabytes the tab will run out of memory. Nothing is metered, rate-limited or held back for a paid tier, because there is no per-use cost on our side to meter. The privacy page says what the server does see, which is page requests and nothing else.

A free account, if you want one

Every tool works signed out, exactly as it always has, and an account gates nothing. What creating one does is remember things between browsers: the tools you star, the ones you used recently, and saved presets per tool, so that the column selection you set up on JSON → CSV is still there on your laptop tomorrow. It stores preferences, never documents — the JSON you paste is never sent anywhere, with or without an account.

FAQ

Is any of this actually free, or is there a paid tier later?

All 99 tools are free and none are limited. The tools run on your computer rather than ours, so there is no per-conversion cost to recover. There is no paid tier, no file-count cap, and no watermark.

Do you see the JSON I paste?

No. There is no endpoint that receives it. The document is read by JavaScript in your tab and never leaves it — you can verify this by opening your browser's network panel while you use a tool, or by disconnecting from the network and watching the tools keep working.

Which JSON Schema draft do these tools use?

Draft 2020-12 by default throughout. Drafts 04, 06, 07 and 2019-09 are selectable on the validator, the generator and the draft migrator, because the differences between them are real and sometimes you are stuck on an old one.

Why are there separate pages for each direction of a conversion?

Because the options that matter are different. Going from JSON to CSV, the question is how to flatten nesting and which columns to keep; coming back, it is which strings to read as numbers, booleans or nulls. One page with a swap button would hide half its controls at any moment.

Is there an API?

Not yet. Everything here is client-side, so there is no service to call — an API would be a different product rather than an endpoint on this one. If you want one, say so; that is how the queue gets ordered.

— S., [email protected]