@zod/core
If you are build a library on top of Zod, you should build against the @zod/core
package. This package exports the core classes that are extended by zod
and @zod/mini
.
The @zod/core
package implements Zod's core parsing logic and additional utilities. It is not intended to be used directly; instead it's designed to be extended by other packages like zod
and @zod/mini
. It implements:
- base schema class
$ZodType
and subclasses like$ZodString
, etc. - base check class
$ZodCheck
and subclasses like$ZodCheckMultipleOf
, etc. - base error class
$ZodError
and shared issue formats - utility types and functions
For simplicity, the zod
and @zod/mini
packages re-export their @zod/core
dependency as z.core
.
Versioning
Though Zod 4 is in beta, the @zod/core
package is published as latest
. The plan is to keep @zod/core
in the It will remain in the v0.x
version range ("initial development" according to semver) until the first stable release of Zod 4, at which point v1.0.0
will be released.
Schemas
The base class for all Zod schemas is $ZodType
. It accepts two generic parameters: Output
and Input
.
@zod/core
exports a number of subclasses that implement some common parsers. A union of all first-party subclasses is exported as z.$ZodTypes
.
All @zod/core
subclasses only contain a single property: _zod
. This property is an object containing the schemas internals: it's state, implemention, etc. The goal is to make @zod/core
as extensible and unopinionated as possible. Other libraries can "build their own Zod" or otherwise integrate Zod's super-optimized parsing logic into their own library by leveraging the subclasses in @zod/core
.
The _zod
internals property contains some notable properties:
.def
— The schema's definition: this is the object you pass into the class's constructor to create an instance. It completely describes the schema, and it's JSON-serializable..def.type
— A string representing the schema's type, e.g."string"
,"object"
,"array"
, etc..def.checks
— An array of checks that are executed by the schema after parsing.
.input
— A virtual property that "stores" the schema's inferred input type..output
— A virtual property that "stores" the schema's inferred output type..run()
— The schema's internal parser implementation.
If you are implementing a tool (say, a code generator) that must traverse Zod schemas, you can cast any schema to $ZodTypes
and use the def
property to discriminate between these classes.
There are a number of subclasses of $ZodString
that implement various string formats. These are exported as z.$ZodStringFormatTypes
.
Checks
Every Zod schema contains an array of checks. These perform post-parsing refinements (and occasionally mutations) that do not affect the inferred type.
The base class for all Zod checks is $ZodCheck
. It accepts a single generic parameter T
.
@zod/core
exports a number of subclasses that perform some common refinements. All first-party subclasses are exported as a union called z.$ZodChecks
.
As with schema types, there are a number of subclasses of $ZodCheckStringFormat
that implement various string formats.
You'll notice some of these string format checks overlap with the string format types above. That's because these classes implement both the $ZodCheck
and $ZodType
interfaces. That is, they can be used as either a check or a type.
Note — If an instance implements $ZodCheck
, it gets preprended to the array of checks (.def.checks
) during parsing.
As with z.$ZodTypes
, you can use the ._zod.def.check
property to discriminate between these classes.
Errors
The base class for all errors in Zod is $ZodError
.
For performance reasons, $ZodError
does not extend the built-in Error
class! So using instanceof Error
will return false
.
- The
zod
package implements a subclass of$ZodError
calledZodError
with some additional convenience methods. - The
@zod/mini
package directly uses$ZodError
Issues
The issues
property corresponds to an array of $ZodIssue
objects. All issues extend the z.$ZodIssueBase
interface.
Zod defines the following issue subtypes:
For details on each type, refer to the implementation.
Best practices
If you're reading this page, you're likely trying to build some kind of tool or library on top of Zod. This section breaks down some best practices for doing so.
- If you're just accepted user-defined schemas, use Standard Schema instead
Zod implements the Standard Schema specification, a standard interface for schema libraries to expose their validation logic and inferred types to third-party tools. If your goal is to accept user-defined schemas, extracting their inferred types, and using them to parse data, then Standard Schema is all you need. Refer to the Standard Schema website/docs for more information.
- Set up
peerDependencies
properly!
If your tool accepts Zod schemas from a consumer/user, you should add "@zod/core"
to peerDependencies
. This lets your users "bring their own Zod". Be as flexible as possible with the version range. For example, if your tool is compatible with @zod/core
, you can use the following. This allows your users to bring any version of @zod/core
, avoiding accidental duplicate installs.
Since package managers generally won't install your own peerDependencies
, you'll need to add @zod/core
to your devDependencies
as well. As new versions of @zod/core
are released, you can update your devDependencies
to match the latest version. This is important for testing and development purposes.
- Allow for new schemas and checks to be added
If you are switch
ing over $ZodTypes
or $ZodChecks
, account for the possibility that new types or checks will be added in future minor versions. The addition of new APIs is not considered a breaking change.
Moreover, some users may define their own custom subclasses of $ZodType
. This is less likely
Consider printing an informative warning to the console and providing some reasonable fallback behavior if you encounter an unknown schema type.