metaschemaFast, zero-dependency schemas for JavaScript

Declare data structures and domain models once as plain objects. Validate and parse them at tens of millions of checks a second, get every problem back as data, and generate TypeScript types and JSON Schema from the same definitions. Node.js and browsers.

Get Started

Why metaschema?

Compare with alternatives

⚡

Fast, with or without eval

Every check is compiled when the schema is built: about 10 million checks a second of a flat object on the closures, about 30 million through the JIT backend, which keeps the closures where new Function is not allowed.

Benchmarks

✍️

Compact schema syntax

Shorthands like '?string', 'key?', { array: 'number' } and ['number', 'number'] sit next to a long form for when a field needs options. A definition is a plain object you can store and send.

Schema syntax

✅

Every problem, as data

check() walks the whole value and returns every issue with a code, the path as keys, its params and a localized message. Collecting all of them is cheap: no exceptions, messages rendered once.

Validation

🔄

Parse, not only a verdict

parse() fills defaults, strips unknown keys, renames aliases and decodes fields into the values of the program; serialize() goes back. The async methods run independent validators together.

Parse and serialize

🔷

TypeScript from the definition

The type of a value a schema accepts, and of what parse gives, is inferred from the definition with nothing generated; a model renders its entities as interfaces.

TypeScript

🔌

Standard Schema and JSON Schema

Every schema is a Standard Schema for tRPC, TanStack Form or Hono. Export JSON Schema 2020-12, draft-07, OpenAPI 3.0, MongoDB or a strict profile for LLM tools, and import JSON Schema back.

JSON Schema

🧩

Domain models that evolve

Entities, registries and projections with references, relations and indexes. A model is a JSON snapshot; diff two of them and classify the change as backward, forward or fully compatible.

Model evolution

🛠️

Tools around the schema

Draw values a schema accepts, infer a definition from data, merge two schemas, validate streams of records, and check email, uuid, date-time and your own string formats.

Sample values

📦

Zero dependencies, any runtime

No runtime dependencies and 21.3 KB min+gzip for the core; generators, the JIT, the model tools, coercion, derive and MongoDB are subpaths you import when you need them. CommonJS with ESM named imports, Node.js 18+ and browsers, no build step.

Browser usage

Benchmarks

A check is a function call, and with the JIT backend a straight-line one. The chart runs the object of the moltar suite through metaschema and ten other validators on one machine, each in a fresh process and checked for the right verdicts first. On the JIT metaschema validates 30 to 48 million objects a second, and in the strict modes (unknown keys rejected) only TypeBox's compiler is faster; on its closures, with no eval at all, 4.6 to 9.8 million, ahead of every interpreter in the set. Asked for every issue of an invalid value with its path, it is second only to Ajv, and 6 to 100 times faster than the other TypeScript-first validators.

Nested object, parseStrict

A valid object of 6 fields and a nested object of 3. parseStrict: unknown keys rejected, the value returned. ops/s, log scale, higher is better.

TypeBoxCompile(…).Check()

43.2M

metaschemacheck() · compile(), JIT

29.9M

Ajvcompile(…)(value), draft 2020-12

23.5M

ZodsafeParse()

7.34M

metaschemacheck() · closures

4.81M

ArkTypetype(value) · allows()

4.02M

ValibotsafeParse() · is()

1.8M

Effect SchemadecodeUnknownResult() · is()

Metarhia metaschemacheck() · strict only

550k

Joivalidate(), convert: false

543k

Superstructvalidate() · is()

266k

YupvalidateSync() · isValidSync(), strict

175k

TypeBoxValue.Check()

118k

Apple M3 Max · 16 cores · Node v24.14.1 · median of 5 × 1 s samples, each library in a fresh process · 2026-10-13

Method, every issue of an invalid value, cold start, bundle size and release over release →

metaschemaFast, zero-dependency schemas for JavaScript

Fast, with or without eval

Compact schema syntax

Every problem, as data

Parse, not only a verdict

TypeScript from the definition

Standard Schema and JSON Schema

Domain models that evolve

Tools around the schema

Zero dependencies, any runtime

Benchmarks ​

Benchmarks