Skip to main content
Version: v3.x

Permissions

Introduction

Access control rules are essential for securing your data and ensuring that only authorized users can access it. In Hasura, these are referred to as permissions. You can use permissions to control access to certain rows or columns in your database, or to restrict access to certain operations or fields in your GraphQL API.

The following types of permissions can be defined:

TypeDescription
TypePermissionsDefine which fields are allowed to be accessed by a role on an output type.
ModelPermissionsDefine which objects or rows within a model are allowed to be accessed by a role.
CommandPermissionsDefine whether the command is executable by a role.

How TypePermissions work

Lifecycle

By default, whenever a new type is created in your supergraph, each field is only accessible to the admin role. You can think of these as permissions on columns in a database table.

You can quickly generate new roles by adding a new item to the permissions array in the TypePermissions object. Each item in the array should have a role field and an output field. The output field should contain an allowedFields array, which lists the fields that are accessible to the role when the type is used in an output context.

To make a new TypePermission or role available in your supergraph, you'll need to create a new build using the CLI.

Examples

---
kind: TypePermissions
version: v1
definition:
typeName: article
permissions:
- role: admin
output:
allowedFields:
- article_id
- author_id
- title
- role: user
output:
allowedFields:
- article_id
- author_id
FieldDescriptionReference
kindIndicates that this object is defining permissions for a specific type.TypePermissions
versionSpecifies the version of the type permissions structure.TypePermissionsV1
definition.typeNameThe name of the type for which permissions are being defined.CustomTypeName
definition.permissionsA list of permissions specifying which fields are accessible for different roles.TypePermission
definition.permissions[].roleThe role for which permissions are being defined.Role
definition.permissions[].output.allowedFieldsThe fields of the type that are accessible for a role when the type is used in an output context.TypeOutputPermission

How ModelPermissions work

Lifecycle

By default, whenever a new model is created in your supergraph, all records are only accessible to the admin role. You can think of these as permissions on rows in a database table.

You can restrict access to certain rows by adding a new item to the permissions array in the ModelPermissions object. Each item in the array should have a role field and a select field. The select field should contain a filter expression that determines which rows are accessible to the role when selecting from the model.

Most commonly, you'll use session variables — provided via your configured authentication mechanism — to restrict access to rows based on the user's identity or other criteria. You can also use argument presets to enforce row-level security. You can see an example of this below.

To make a new ModelPermission or role available in your supergraph, after updating your metadata, you'll need to create a new build using the CLI.

Examples

---
kind: ModelPermissions
version: v1
definition:
modelName: Articles
permissions:
- role: admin
select:
filter: null
- role: user
select:
filter:
fieldComparison:
field: author_id
operator: _eq
value:
sessionVariable: x-hasura-user-id
FieldDescriptionReference
kindIndicates that this object is defining permissions for a specific model.ModelPermissions
versionSpecifies the version of the model permissions structure.ModelPermissionsV1
definition.modelNameThe name of the model for which permissions are being defined.ModelName
definition.permissionsA list of permissions specifying which rows or objects are accessible for different roles.ModelPermission
definition.permissions[].roleThe role for which permissions are being defined.Role
definition.permissions[].select.filterA filter expression that determines which rows are accessible for this role when selecting from the model.ModelPredicate

How CommandPermissions work

Lifecycle

By default, whenever a new command is created in your supergraph, it is only executable by the admin role.

You can enable or restrict access to commands by adding a new item to the permissions array in the CommandPermissions object. Each item in the array should have a role field and an allowExecution field. The allowExecution field should be set to true if the command is executable by the role.

Further, you can use argument presets to pass actual logical expressions to your data sources to control how they do things.

For example, a data connector might expose a Command called delete_user_by_id with two arguments - user_id and pre_check. user_id is the primary key of the user you'd like to remove, and pre_check lets you provide a custom boolean expression.

kind: CommandPermissions
version: v1
definition:
commandName: delete_user_by_id
permissions:
- role: admin
allowExecution: true
- role: user
allowExecution: true
argumentPresets:
- argument: pre_check
value:
booleanExpression:
fieldComparison:
field: is_invincible
operator: _eq
value:
literal: false

Now, when admin runs this command, once again, they can do what they want, and provide their own pre_check if they want. The user however, is able to pass a user_id argument, but the pre_check expression is passed to the data connector which will only let them delete the row if the row's is_invincible value is set to false.

To make a execution of a command available to a role in your supergraph, after updating your metadata, you'll need to create a new build using the CLI.

Examples

---
kind: CommandPermissions
version: v1
definition:
commandName: get_article_by_id
permissions:
- role: admin
allowExecution: true
- role: user
allowExecution: true
argumentPresets:
- argument: id
value:
literal: 100
FieldDescriptionReference
kindIndicates that this object is defining permissions for a specific command.CommandPermissions
versionSpecifies the version of the command permissions structure.CommandPermissionsV1
definition.commandNameThe name of the command for which permissions are being defined.CommandName
definition.permissionsA list of permissions specifying whether the command is executable by different roles.CommandPermission
definition.permissions[].roleThe role for which permissions are being defined.Role
definition.permissions[].allowExecutionIndicates whether the command is executable by the role.CommandPermission
definition.permissions[].argumentPresetsPreset values for arguments that are enforced when the command is executed by the role.ArgumentPreset

Metadata structure

TypePermissions

Definition of permissions for an OpenDD type.

KeyValueRequiredDescription
kindTypePermissionstrue
versionv1true
definitionTypePermissionsV1trueDefinition of permissions for an OpenDD type.

Example:

kind: TypePermissions
version: v1
definition:
typeName: article
permissions:
- role: admin
output:
allowedFields:
- article_id
- author_id
- title
- role: user
output:
allowedFields:
- article_id
- author_id

TypePermissionsV1

Definition of permissions for an OpenDD type.

KeyValueRequiredDescription
typeNameCustomTypeNametrueThe name of the type for which permissions are being defined. Must be an object type.
permissions[TypePermission]trueA list of type permissions, one for each role.

TypePermission

Defines permissions for a particular role for a type.

KeyValueRequiredDescription
roleRoletrueThe role for which permissions are being defined.
outputTypeOutputPermission / nullfalsePermissions for this role when this type is used in an output context. If null, this type is inaccessible for this role in an output context.
inputTypeInputPermission / nullfalsePermissions for this role when this type is used in an input context. If null, this type is accessible for this role in an input context.

Example:

role: user
output:
allowedFields:
- article_id
- author_id
input:
fieldPresets:
- field: author_id
value:
sessionVariable: x-hasura-user-id

TypeInputPermission

Permissions for a type for a particular role when used in an input context.

KeyValueRequiredDescription
fieldPresets[FieldPreset]falsePreset values for fields of the type

FieldPreset

Preset value for a field

KeyValueRequiredDescription
fieldFieldNametrueField name for preset
valueValueExpressiontrueValue for preset

ValueExpression

An expression which evaluates to a value that can be used in permissions and various presets.

Must have exactly one of the following fields:

KeyValueRequiredDescription
literalfalse
sessionVariablestringfalse

TypeOutputPermission

Permissions for a type for a particular role when used in an output context.

KeyValueRequiredDescription
allowedFields[FieldName]trueFields of the type that are accessible for a role

FieldName

The name of a field in a user-defined object type.

Value: string

Role

The role for which permissions are being defined.

Value: string

CustomTypeName

The name of a user-defined type.

Value: string

ModelPermissions

Definition of permissions for an OpenDD model.

KeyValueRequiredDescription
kindModelPermissionstrue
versionv1true
definitionModelPermissionsV1trueDefinition of permissions for an OpenDD model.

Examples:

kind: ModelPermissions
version: v1
definition:
modelName: Articles
permissions:
- role: admin
select:
filter: null
- role: user
select:
filter:
fieldComparison:
field: author_id
operator: _eq
value:
sessionVariable: x-hasura-user-id
kind: ModelPermissions
version: v1
definition:
modelName: Articles
permissions:
- role: admin
select:
filter: null
- role: user
select:
filter:
relationship:
name: author
predicate:
fieldComparison:
field: id
operator: _eq
value:
sessionVariable: x-hasura-user-id

ModelPermissionsV1

Definition of permissions for an OpenDD model.

KeyValueRequiredDescription
modelNameModelNametrueThe name of the model for which permissions are being defined.
permissions[ModelPermission]trueA list of model permissions, one for each role.

ModelPermission

Defines the permissions for an OpenDD model.

KeyValueRequiredDescription
roleRoletrueThe role for which permissions are being defined.
selectSelectPermission / nullfalseThe permissions for selecting from this model for this role. If this is null, the role is not allowed to query the model.

Example:

role: user
select:
filter:
fieldComparison:
field: author_id
operator: _eq
value:
sessionVariable: x-hasura-user-id
argument_presets:
- field: likes_dogs
value:
literal: true

SelectPermission

Defines the permissions for selecting a model for a role.

KeyValueRequiredDescription
filternull / ModelPredicatetrueFilter expression when selecting rows for this model. Null filter implies all rows are selectable.
argumentPresets[ArgumentPreset]falsePreset values for arguments for this role
allowSubscriptionsbooleanfalseWhether the role is allowed to subscribe to the root fields of this model.

ArgumentPreset

Preset value for an argument

KeyValueRequiredDescription
argumentArgumentNametrueArgument name for preset
valueValueExpressionOrPredicatetrueValue for preset

ValueExpressionOrPredicate

An expression which evaluates to a value that can be used in permissions and various presets.

Must have exactly one of the following fields:

KeyValueRequiredDescription
literalfalse
sessionVariablestringfalseUsed to represent the name of a session variable, like "x-hasura-role".
booleanExpressionModelPredicatefalse

ArgumentName

The name of an argument.

Value: string

ModelPredicate

A predicate that can be used to restrict the objects returned when querying a model.

Must have exactly one of the following fields:

KeyValueRequiredDescription
fieldComparisonFieldComparisonPredicatefalseField comparison predicate filters objects based on a field value.
fieldIsNullFieldIsNullPredicatefalsePredicate to check if the given field is null.
relationshipRelationshipPredicatefalseRelationship predicate filters objects of a source model based on a predicate on the related model.
and[ModelPredicate]false
or[ModelPredicate]false
notModelPredicatefalse

Examples:

fieldComparison:
field: author_id
operator: _eq
value:
sessionVariable: x-hasura-user-id
relationship:
name: author
predicate:
fieldComparison:
field: id
operator: _eq
value:
sessionVariable: x-hasura-user-id
and:
- fieldComparison:
field: author_id
operator: _eq
value:
sessionVariable: x-hasura-user-id
- fieldComparison:
field: title
operator: _eq
value:
literal: Hello World
not:
fieldComparison:
field: author_id
operator: _eq
value:
sessionVariable: x-hasura-user-id

RelationshipPredicate

Relationship predicate filters objects of a source model based on a predicate on the related model.

KeyValueRequiredDescription
nameRelationshipNametrueThe name of the relationship of the object type of the model to follow.
predicateModelPredicate / nullfalseThe predicate to apply on the related objects. If this is null, then the predicate evaluates to true as long as there is at least one related object present.

RelationshipName

The name of the GraphQL relationship field.

Value: string

FieldIsNullPredicate

Predicate to check if the given field is null.

KeyValueRequiredDescription
fieldFieldNametrueThe name of the field that should be checked for a null value.

FieldComparisonPredicate

Field comparison predicate filters objects based on a field value.

KeyValueRequiredDescription
fieldFieldNametrueThe field name of the object type of the model to compare.
operatorOperatorNametrueThe name of the operator to use for comparison.
valueValueExpressiontrueThe value expression to compare against.

ValueExpression

An expression which evaluates to a value that can be used in permissions and various presets.

Must have exactly one of the following fields:

KeyValueRequiredDescription
literalfalse
sessionVariablestringfalse

OperatorName

The name of an operator

Value: string

FieldName

The name of a field in a user-defined object type.

Value: string

Role

The role for which permissions are being defined.

Value: string

ModelName

The name of data model.

Value: string

CommandPermissions

Definition of permissions for an OpenDD command.

KeyValueRequiredDescription
kindCommandPermissionstrue
versionv1true
definitionCommandPermissionsV1trueDefinition of permissions for an OpenDD command.

Example:

kind: CommandPermissions
version: v1
definition:
commandName: get_article_by_id
permissions:
- role: admin
allowExecution: true
- role: user
allowExecution: true

CommandPermissionsV1

Definition of permissions for an OpenDD command.

KeyValueRequiredDescription
commandNameCommandNametrueThe name of the command for which permissions are being defined.
permissions[CommandPermission]trueA list of command permissions, one for each role.

CommandPermission

Defines the permissions for a role for a command.

KeyValueRequiredDescription
roleRoletrueThe role for which permissions are being defined.
allowExecutionbooleantrueWhether the command is executable by the role.
argumentPresets[ArgumentPreset]falsePreset values for arguments for this role

Example:

role: user
allowExecution: true
argumentPresets:
- argument: user_id
value:
session_variable: x-hasura-user_id

ArgumentPreset

Preset value for an argument

KeyValueRequiredDescription
argumentArgumentNametrueArgument name for preset
valueValueExpressionOrPredicatetrueValue for preset

ValueExpressionOrPredicate

An expression which evaluates to a value that can be used in permissions and various presets.

Must have exactly one of the following fields:

KeyValueRequiredDescription
literalfalse
sessionVariablestringfalseUsed to represent the name of a session variable, like "x-hasura-role".
booleanExpressionModelPredicatefalse

ModelPredicate

A predicate that can be used to restrict the objects returned when querying a model.

Must have exactly one of the following fields:

KeyValueRequiredDescription
fieldComparisonFieldComparisonPredicatefalseField comparison predicate filters objects based on a field value.
fieldIsNullFieldIsNullPredicatefalsePredicate to check if the given field is null.
relationshipRelationshipPredicatefalseRelationship predicate filters objects of a source model based on a predicate on the related model.
and[ModelPredicate]false
or[ModelPredicate]false
notModelPredicatefalse

Examples:

fieldComparison:
field: author_id
operator: _eq
value:
sessionVariable: x-hasura-user-id
relationship:
name: author
predicate:
fieldComparison:
field: id
operator: _eq
value:
sessionVariable: x-hasura-user-id
and:
- fieldComparison:
field: author_id
operator: _eq
value:
sessionVariable: x-hasura-user-id
- fieldComparison:
field: title
operator: _eq
value:
literal: Hello World
not:
fieldComparison:
field: author_id
operator: _eq
value:
sessionVariable: x-hasura-user-id

RelationshipPredicate

Relationship predicate filters objects of a source model based on a predicate on the related model.

KeyValueRequiredDescription
nameRelationshipNametrueThe name of the relationship of the object type of the model to follow.
predicateModelPredicate / nullfalseThe predicate to apply on the related objects. If this is null, then the predicate evaluates to true as long as there is at least one related object present.

RelationshipName

The name of the GraphQL relationship field.

Value: string

FieldIsNullPredicate

Predicate to check if the given field is null.

KeyValueRequiredDescription
fieldFieldNametrueThe name of the field that should be checked for a null value.

FieldComparisonPredicate

Field comparison predicate filters objects based on a field value.

KeyValueRequiredDescription
fieldFieldNametrueThe field name of the object type of the model to compare.
operatorOperatorNametrueThe name of the operator to use for comparison.
valueValueExpressiontrueThe value expression to compare against.

ValueExpression

An expression which evaluates to a value that can be used in permissions and various presets.

Must have exactly one of the following fields:

KeyValueRequiredDescription
literalfalse
sessionVariablestringfalse

OperatorName

The name of an operator

Value: string

FieldName

The name of a field in a user-defined object type.

Value: string

ArgumentName

The name of an argument.

Value: string

Role

The role for which permissions are being defined.

Value: string

CommandName

The name of a command.

Value: string