JSON → TypeScript

What is JSON?

JSON (JavaScript Object Notation) is a lightweight, text-based format for representing structured data. It was derived from JavaScript object syntax but is language-independent: every major language can read and write it. Its job is to move structured data between systems, API responses, config files, message payloads, in a form that is both human-readable and trivially machine-parseable. That combination is why it became the default data format of the modern web.

JSON data types

JSON allows exactly six value types, no more:

• string (double-quoted), number, boolean (true/false), and null.
• array (an ordered list in [ ]) and object (key/value pairs in { }, keys must be strings).

Notably absent: dates (use ISO 8601 strings), comments, trailing commas, undefined, and functions. Those are the source of most “why won’t my JSON parse” bugs (see common mistakes).

JSON in code

Every language has a parse step (text to value) and a serialize step (value to text). JavaScript:

// parse a JSON string into an object
const obj = JSON.parse('{"id":42,"name":"Flat White"}');

// serialize an object back to JSON (the 2 is the pretty-print indent)
const text = JSON.stringify(obj, null, 2);

Python:

import json

# parse a JSON string into a dict
data = json.loads('{"id": 42, "name": "Flat White"}')

# serialize a dict back to JSON
text = json.dumps(data, indent=2)

Want types for that data? That is exactly what the tool above does, paste the JSON and get TypeScript interfaces.

JSON in AI

JSON quietly became the language AI systems speak to the rest of your code. It did not exist as a concern when most JSON tutorials were written, but in 2026 it is everywhere a model meets software:

• Structured outputs: models return JSON that matches a schema you define, so the result drops straight into your program. See structured outputs and build one with the Structured Output Builder.
• Tool / function calling: when a model calls a tool, its arguments come back as JSON, see tool calling.
• MCP: the Model Context Protocol is built on JSON-RPC.
• API responses: the data you feed a model, and get back, is almost always JSON.

JSON vs XML

Common mistakes

• Trailing commas. Valid in JavaScript, invalid in JSON. [1, 2,] will not parse.
• Single quotes. JSON strings and keys must use double quotes.
• Comments. Not allowed in standard JSON; strip them first.
• Unquoted keys. { name: "x" } is a JS object, not JSON; keys must be quoted.
• Wrong types for dates / big numbers. Use strings; very large integers can lose precision as JSON numbers.

About this tool

Paste a JSON object or array on the left and KB Cafe walks the structure to produce matching interface declarations on the right, entirely in your browser. The inference rules:

• Objects become an interface; nested objects get a named child interface from the key.
• Arrays are typed by merging every element, [1, 2] is number[], and an array of objects merges their keys (missing keys become optional).
• null values widen a field to … | null rather than guessing.
• Empty arrays / objects fall back to unknown[] and Record<string, unknown> so nothing is silently wrong.

FAQ

Related

JSON → Go · CSV → JSON · YAML → JSON · JWT decoder (Base64URL JSON) · Base64 · HTTP · Structured Output Builder.