TypeScript-first schema declaration and validation library with static type inference
23 Aug 2025 · Newest first
Last checked
Checking sources…
Showing 1 of 1 releases. Full release notes.
The first minor version since the introduction of Zod 4 back in May. This version contains a number of features that barely missed the cut for the 4.0 release. With Zod 4 stable and widely adopted, there's more time to resume feature development.
This is the flagship feature of this release. Codecs are a new API & schema type that encapsulates a bi-directional transformation. It's a huge missing piece in Zod that's finally filled, and it unlocks some totally new ways to use Zod.
Signals highlight changes worth reviewing. They cannot determine whether your application is affected.
const stringToDate = z.codec(
z.iso.datetime(), // input schema: ISO date string
z.date(), // output schema: Date object
{
decode: (isoString) => new Date(isoString),
encode: (date) => date.toISOString(),
}
);
New top-level functions are added for processing inputs in the forward direction ("decoding") and backward direction ("encoding").
stringToDate.decode("2025-08-21T20:59:45.500Z")
// => Date
stringToDate.encode(new Date())
// => "2025-08-21T20:59:45.500Z"
Note — For bundle size reasons, these new methods have not added to Zod Mini schemas. Instead, this functionality is available via equivalent top-level functions.
// equivalent at runtime z.decode(stringToDate, "2024-01-15T10:30:00.000Z"); z.encode(stringToDate, new Date());
.parse() vs .decode()Both .parse() and decode() process data in the "forward" direction. They behave identically at runtime.
stringToDate.parse("2025-08-21T20:59:45.500Z");
stringToDate.decode("2025-08-21T20:59:45.500Z");
There is an important difference however. While .parse() accepts any input, .decode() expects a strongly typed input. That is, it expects an input of type string, whereas .parse() accepts unknown.
stringToDate.parse(Symbol('not-a-string'));
// => fails at runtime, but no TypeScript error
stringToDate.decode(Symbol("not-a-string"));
// ^ ❌ Argument of type 'symbol' is not assignable to parameter of type 'Date'. ts(2345)
You can use any Zod schema with .encode(). The vast majority of Zod schemas are non-transforming (the input and output types are identical) so .decode() and .encode() behave identically. Only certain schema types change their behavior:
B->A and executes the encode transform during encodingB->A instead of A->BNote — To avoid increasing bundle size unnecessarily, these new methods are not available on Zod Mini schemas. For those schemas, equivalent top-level functions are provided.
The usual async and safe variants exist as well:
// decode methods
stringToDate.decode("2024-01-15T10:30:00.000Z")
await stringToDate.decodeAsync("2024-01-15T10:30:00.000Z")
stringToDate.safeDecode("2024-01-15T10:30:00.000Z")
await stringToDate.safeDecodeAsync("2024-01-15T10:30:00.000Z")
// encode methods
stringToDate.encode(new Date())
await stringToDate.encodeAsync(new Date())
stringToDate.safeEncode(new Date())
await stringToDate.safeEncodeAsync(new Date())
Below are some "worked examples" for some commonly-needed codecs. These examples are all tested internally for correctness. Just copy/paste them into your project as needed. There is a more comprehensive set available at zod.dev/codecs.
stringToBigIntConverts bigint into a serializable form.
const stringToBigInt = z.codec(z.string(), z.bigint(), {
decode: (str) => BigInt(str),
encode: (bigint) => bigint.toString(),
});
stringToBigInt.decode("12345"); // => 12345n
stringToBigInt.encode(12345n); // => "12345"
jsonParses/stringifies JSON data.
const jsonCodec = z.codec(z.string(), z.json(), {
decode: (jsonString, ctx) => {
try {
return JSON.parse(jsonString);
} catch (err: any) {
ctx.issues.push({
code: "invalid_format",
format: "json_string",
input: jsonString,
message: err.message,
});
return z.NEVER;
}
},
encode: (value) => JSON.stringify(value),
});
To further validate the data, .pipe() the result of this codec into another schema.
const Params = z.object({ name: z.string(), age: z.number() });
const JsonToParams = jsonCodec.pipe(Params);
JsonToParams.decode('{"name":"Alice","age":30}'); // => { name: "Alice", age: 30 }
JsonToParams.encode({ name: "Bob", age: 25 }); // => '{"name":"Bob","age":25}'
For more examples and a technical breakdown of how encoding works, reads theannouncement blog post and new Codecs docs page. The docs page contains implementations for several other commonly-needed codecs:
.safeExtend()The existing way to add additional fields to an object is to use .extend().
const A = z.object({ a: z.string() })
const B = A.extend({ b: z.string() })
Unfortunately this is a bit of a misnomer, as it allows you to overwrite existing fields. This means the result of .extend() may not literally extend the original type (in the TypeScript sense).
const A = z.object({ a: z.string() }) // { a: string }
const B = A.extend({ a: z.number() }) // { a: number }
To enforce true extends logic, Zod 4.1 introduces a new .safeExtend() method. This statically enforces that the newly added properties conform to the existing ones.
z.object({ a: z.string() }).safeExtend({ a: z.number().min(5) }); // ✅
z.object({ a: z.string() }).safeExtend({ a: z.any() }); // ✅
z.object({ a: z.string() }).safeExtend({ a: z.number() });
// ^ ❌ ZodNumber is not assignable
Importantly, this new API allows you to safely extend objects containing refinements.
const AB = z.object({ a: z.string(), b: z.string() }).refine(val => val.a === val.b);
const ABC = AB.safeExtend({ c: z.string() });
// ABC includes the refinements defined on AB
Previously (in Zod 4.x) any refinements attached to the base schema were dropped in the extended result. This was too unexpected. It now throws an error. (Zod 3 did not support extension of refined objects either.)
z.hash()A new top-level string format for validating hashes produced using various common algorithms & encodings.
const md5Schema = z.hash("md5");
// => ZodCustomStringFormat<"md5_hex">
const sha256Base64 = z.hash("sha256", { enc: "base64" });
// => ZodCustomStringFormat<"sha256_base64">
The following hash algorithms and encodings are supported. Each cell provides information about the expected number of characters/padding.
| Algorithm / Encoding | "hex" | "base64" | "base64url" |
|---|---|---|---|
"md5" | 32 | 24 (22 + "==") | 22 |
"sha1" | 40 | 28 (27 + "=") | 27 |
"sha256" | 64 | 44 (43 + "=") | 43 |
"sha384" | 96 | 64 (no padding) | 64 |
"sha512" | 128 | 88 (86 + "==") | 86 |
z.hex()To validate hexadecimal strings of any length.
const hexSchema = z.hex();
hexSchema.parse("123abc"); // ✅ "123abc"
hexSchema.parse("DEADBEEF"); // ✅ "DEADBEEF"
hexSchema.parse("xyz"); // ❌ ZodError
FFFFFFFF-FFFF-FFFF-FFFF-FFFFFFFFFFFF) per the RFC$ZodFunction is now a subtype of $ZodType