TOON to JSON converter
The way back. Paste TOON, whether it came out of our converter or out of a model, and get indented JSON with every tabular row expanded into a full object. The parser is deliberately strict: a row count that disagrees with its header is an error, not a guess. Nothing you paste leaves your browser.
Text is processed by your own browser and never leaves your device: this page has no server to send it to.
How it works
Lines are read by indentation. A header like items[2]{id,name}: takes the next two indented rows and rebuilds them into objects using the declared field names, checking both the count and the cells per row. Quoted values are unescaped, unquoted ones are typed the way the spec says: true, false and null become their literals, numeric text becomes a number, everything else stays a string. Built against the published TOON specification, checked 2026-08-23.
Frequently asked questions
Why does it refuse a document a model produced?
Usually a count mismatch: the model wrote items[3] and then listed two rows, or a row is missing a cell. The error names the line so you can fix it. That strictness is on purpose, because silently accepting a short row would hand you JSON with a field quietly missing, which is worse than an error.
Does a JSON to TOON to JSON round trip give me back the same thing?
For the constructs this converter supports, yes, including the awkward cases: a string "42", an empty string, a value containing a comma, and keys that need quoting all survive both directions. Key order is preserved too. What does not survive is anything JSON itself loses, such as undefined or a Date object, which never made it into the JSON in the first place.
What about the tab and pipe variants?
Not supported. The spec allows tab or pipe delimiters and a keyed tabular form; this converter implements the comma-delimited core only, and will error rather than misread one of the others. If you need those, say so and we will look at the demand.