wellformed

Introduction

Portable validation schemas for TypeScript and Rust. Author once, validate everywhere.

Validation logic should be data. Write a schema once in TypeScript. It compiles to JSON. The same rules run in TypeScript and Rust, identically. No backend rewrite. No drift.

TypeScript builder -> JSON IR -> TypeScript runtime
                              -> Rust runtime
npm install wellformed-ts
pnpm add wellformed-ts
yarn add wellformed-ts
bun add wellformed-ts
cargo add wellformed wellformed-macros serde_json

Your first schema, in 30 seconds

import { w, validate, type Infer } from "wellformed-ts";

const Contact = w.object({
  firstName: w.string().trim().minLen(1),
  lastName: w.string().trim().minLen(1),
  email: w.string().trim().lower().email(),
  phone: w.string().phoneUS().optional(),
  preferredContact: w.enum(["email", "phone", "mail"] as const),
});

// Types come free. No duplicate definitions.
type Contact = Infer<typeof Contact>;
// { firstName: string; lastName: string; email: string;
//   phone?: string; preferredContact: "email" | "phone" | "mail" }

const result = validate(Contact.toSchema("1.0"), input);
result.valid; // true | false
use serde_json::json;
use wellformed_macros::wellformed;

fn main() -> wellformed::Result<()> {
    let contact = wellformed!("schemas/contact.json");
    let (result, value) = contact.validate_json(r#"{
      "firstName": "  Ada  ",
      "lastName": "Lovelace",
      "email": "ADA@EXAMPLE.COM",
      "preferredContact": "email"
    }"#)?;

    assert!(result.is_valid());
    assert_eq!(value["firstName"], json!("Ada"));

    Ok(())
}

TypeScript gives you the fluent authoring surface and inferred application type. Rust uses wellformed! to embed the same JSON IR as a schema value with native validation helpers.

Run this in the Playground

Why this is different

Zod hides your rules inside JavaScript functions. wellformed keeps them as JSON. That single decision is the whole product: the schema becomes portable.

import { w, schemaToJSON } from "wellformed-ts";

const schema = w
  .object({
    ssn: w.string().digitsOnly().ssn(),
  })
  .toSchema("1.0");

schemaToJSON(schema, true);
{
  "version": "1.0",
  "root": {
    "type": "object",
    "properties": {
      "ssn": {
        "type": "string",
        "transforms": [{ "fn": "digits_only" }],
        "constraints": [
          {
            "pred": { "type": "call", "name": "is_ssn" },
            "error": {
              "code": "INVALID_SSN",
              "message": "Invalid Social Security Number"
            }
          }
        ]
      }
    }
  }
}

Every transform, predicate, and error code is plain JSON. Store it in a database. Send it over the wire. Deserialize it in the Rust wellformed crate. Same results, no JavaScript required.

The whole idea, in one sentence

Validation that crosses a runtime boundary should be data, not a closure.

Where to go next

Getting Started

Install and validate your first schema in under a minute.

Playground

Edit a schema and watch a real form validate live.

Validate in Rust

Run the same IR natively in the Rust crate.

Domain validators

60+ built-in predicates: SSN, EIN, CUSIP, IBAN, ICD-10.

Cross-field rules

Conditional requirements and comparisons, as data.

vs Zod and JSON Schema

When to reach for wellformed, and when not to.

Getting Started

Install wellformed and validate your first schema in under a minute.

On this page

Your first schema, in 30 secondsWhy this is differentWhere to go next