JSON to Python — generate Python dataclasses online

Have a JSON response or config file and want real Python types for it instead of passing dicts around? Generate standard-library @dataclass definitions in one click, entirely in your browser.

Json Web Viewer's Types mode reads your document's actual shape — key names, nesting, array element types, which fields are sometimes missing or null — and generates matching Python dataclasses. The output uses only the standard library (dataclasses and typing), so it runs with no pip install. No server round-trip, no account, no upload.

Field names are converted to Python's snake_case, with a comment noting the original JSON key wherever the name changed. Nested objects become their own dataclasses, declared below the root — from __future__ import annotations is emitted so those forward references resolve. And a field that a dataclass rule would otherwise break on — a default-valued field before a required one is a TypeError at import time — is ordered correctly, not just prettily.

Example — this JSON:

{
  "id": "ord_9f21a",
  "customer": {
    "name": "Ada Lovelace",
    "email": "ada@example.com",
    "vip": true
  },
  "items": [
    { "sku": "book-001", "qty": 2, "price": 19.99 },
    { "sku": "mug-014", "qty": 1, "price": 12.5 }
  ],
  "shippedAt": null
}

generates this Python:

from __future__ import annotations

from dataclasses import dataclass


@dataclass
class Root:
    customer: Customer
    id: str
    items: list[ItemsItem]
    shipped_at: None  # JSON key: "shippedAt"


@dataclass
class Customer:
    email: str
    name: str
    vip: bool


@dataclass
class ItemsItem:
    price: float
    qty: int
    sku: str

Open this example in the editor →
The link above pre-loads the order example into Types mode. TypeScript is the default target — select Python from the language dropdown next to the mode buttons to regenerate the same document as dataclasses.

Frequently asked questions

Does generating Python dataclasses upload my JSON anywhere?

No. Type generation runs entirely in your browser — your JSON never leaves the page.

Why dataclasses and not Pydantic models or a TypedDict?

A dataclass is standard library, so the generated file runs with no third-party dependency, and it carries real field metadata (defaults, ordering) that a TypedDict cannot express. Pydantic is a better validator, but it splits into incompatible v1/v2 syntaxes and forces a dependency — this tool generates a type, not a validator.

How are missing and null fields handled?

They're kept distinct, because Python can express both: a key absent from some records becomes Optional[T] = None (constructing without it is legal), while a key that's always present but sometimes null becomes Optional[T] with no default — still required, may be None.

Need a different target language? The same dropdown also generates TypeScript, Java, C#, and Kotlin — plus Go and Rust.