• Docs
  • Rules


Rules let you define reusable blocks of permissions that you use in Shield. GraphQL Shield packs a collection of general rules that make your permissions easier to write and your rules more reusable. Each rule you create should be easily explicable in one sentantce and should check exactly one permission.

For example, it's a good practice to have a rule that checks if someone is an admin and another one that checks if someone owns some data, and a bad practice to have both checks in one rule.

Basic rules

allow and deny are GraphQL Shield predefined rules.

allow and deny rules do exactly what their names describe.

Creating your own rules

To create a new rule, wrap your validation function in rule wrapper like in the example below. Your function can take use any parameter from the regular GraphQL resolver function and should return a boolean.

const isAdmin = rule()(async (parent, args, ctx, info) => {
  return ctx.user.isAdmin

Input Rules

Validate arguments using Yup.

function inputRule(name?: string)((yup: Yup, ctx: any) => Yup.Schema, options?: Yup.ValidationOptions): Rule

Input rule works exactly as any other rule would work. Instead of providing a complex validation rule you can simply provide a Yup validation schema which will be mached against provided arguments. This can be especially useful when limiting optional fields such as create and connect with Prisma, for example.


type Mutation {
  login(email: String): LoginPayload

Note that Yup receives entire args object, therefore, you should start composing schema with an object.

const isEmailEmail = inputRule()(
  (yup) =>
      email: yup.string().email('It has to be an email!').required(),
  { abortEarly: false },

Logic Rules

Logic rules let you combine multiple rules into a composition. This should be your glue for making more complex permission requirements out of simple rules.

  • and - allows access only if all subrules used return true
  • chain - rule allows you to chain the rules, meaning that rules won't be executed all at once, but one by one until one fails or all pass
  • or - resolves to true if at least one rule passes
  • race - rule allows you to chain the rules so that execution stops once one of them returns true

Think of race and chain as lazy versions of and and or.

  • not works as usual not in code works.
You may also add a custom error message as the second parameter to not(rule, error).

Here's an example:

import { shield, rule, and, or } from 'graphql-shield'
const isAdmin = rule()(async (parent, args, ctx, info) => {
  return ctx.user.role === 'admin'
const isEditor = rule()(async (parent, args, ctx, info) => {
  return ctx.user.role === 'editor'
const isOwner = rule()(async (parent, args, ctx, info) => {
  return ctx.user.items.some((id) => id === parent.id)
const permissions = shield({
  Query: {
    users: or(isAdmin, isEditor),
  Mutation: {
    createBlogPost: or(isAdmin, and(isOwner, isEditor)),
  User: {
    secret: isOwner,


GraphQL Shield uses a caching mechanism to evaluate rules more efficiently. Since rules are evaluated on per-field basis, this is a crucial component that makes graphql-shield fast. You can choose from three different caching options:

  1. no_cache - prevents rules from being cached.
  2. contextual - use when rule only relies on context parameter (useful for authentication).
  3. strict - use when rule relies on parent or args parameter as well (field specific modifications).
// Contextual
const isAdmin = rule({ cache: 'contextual' })(async (parent, args, ctx, info) => {
  return ctx.user.isAdmin
// Strict
const canSeeUserSensitiveData = rule({ cache: 'strict' })(async (parent, args, ctx, info) => {
  /* The id of observed User matches the id of authenticated viewer. */
  return ctx.viewer.id === parent.id
// No-cache (default)
const admin = rule({ cache: 'no_cache' })(async (parent, args, ctx, info) => {
  return ctx.user.isAdmin || args.code === 'secret' || parent.id === 'theone'


Due to design decisions of how GraphQL Shield works, there are some limitations to rule names and functions:

  • All rules must have a distinct name. Usually, you won't have to care about this as all names are by default automatically generated to prevent such problems. In case your function needs additional variables from other parts of the code and is defined as a function, you'll set a specific name to your rule to avoid name generation.
// Normal
const admin = rule({ cache: 'contextual' })(async (parent, args, ctx, info) => true)
// With external data
const admin = (bool) => rule(`name-${bool}`, { cache: 'contextual' })(async (parent, args, ctx, info) => bool)
  • Cache is disabled by default. To enable cache generation, set cache option when generating a rule.

Backward compatibility: { cache: false } converts to no_cache, and { cache: true } converts to strict.

Last updated on October 10, 2022