Build a Serverless GraphQL API on AWS [Tutorial]

4 min read -

Serverless is a scalable, fast and reliable architecture for APIs. By deploying your API to cloud providers such as AWS, you don't have to think about server administration, you just focus on code. Using the Serverless framework, it becomes really easy as you can deploy a Lambda function in seconds.

GraphQL is a query language simplifying client-server interactions. It's also very efficient thanks to its declarative data fetching. It's no surprise that GitHub, Coursera or AirBnB are using it.

Today we're going to bring the best of both worlds, by building a Serverless GraphQL API and deploy it to AWS !

Here are the steps:

  • Setup the Serverless framework
  • Test the Lambda locally
  • Setup GraphQL with Apollo Server
  • Manage data through DynamoDB
  • Deploy the Lambda to AWS

The full code of this tutorial is available on GitHub, feel free to use it for your next project ! It's a more complete version, with TypeScript, ESLint and Prettier.


To use the Serverless framework and deploy to AWS, you have to do the following:

Setup the Serverless framework

First, init your Node.JS project:

yarn init

Then, add the Serverless configuration in serverless.yml:

service: serverless-api-example # service name

  name: aws # deploy to AWS
  runtime: nodejs12.x
  versionFunctions: false # disable function versioning
  region: us-east-1 # AWS region
    shouldStartNameWithService: true

    handler: src/graphql.handler # Lambda handler path
      - http:
          path: graphql # define Lambda endpoint
          method: any # any HTTP method can access this Lambda
          cors: true # enable CORS
The full list of options is in the Serverless framework documentation

We gave Serverless instructions to deploy our new service serverless-api-example. It has a Lambda called graphql and is accessible at /graphql through HTTP requests.

Let's create the handler for our Lambda, for the moment put the following in src/graphql.js:

exports.handler = () => {
  console.log('wow such api')

Test a Serverless function in local

I recommend you to install the serverless-offline plugin to test your Lambda locally:

yarn add -D serverless-offline

Add this at the end of serverless.yml:

  - serverless-offline

And this script in package.json:

  "scripts": {
    "start": "sls offline"

Finally, enter yarn start and go to localhost:3000/dev/graphql:

> yarn start

offline: [HTTP] server ready: http://localhost:3000 🚀
offline: ANY /dev/graphql (λ: graphql)

# When going to http://localhost:3000/dev/graphql
wow such api # It works ! 🎉

Now that we created our Lambda, let's code its behavior !

Setup a GraphQL Lambda with Apollo Server

apollo-server-lambda is a great way to build a GraphQL API in Serverless. Add it to your project:

yarn add -D apollo-server-lambda

Define the GraphQL schema in src/schema.graphql:

type Query {
  item(id: ID!): Item

type Mutation {
  createItem(content: String!): Item

type Item {
  id: ID!
  content: String

In the schema, We're defining what's an Item: an entity containing id and content. We also define a mutation createItem, which creates an item from a content, and a query item which gets it from an id.

Now, add the following in src/graphql.js:

const fs = require('fs')
const { ApolloServer, gql } = require('apollo-server-lambda')

// get the GraphQL schema
const schema = fs.readFileSync('./src/schema.graphql', 'utf8')

// resolver functions
const resolvers = { 
  Query: {
    item: () => {},

  Mutation: {
    createItem: () => {},

const server = new ApolloServer({ typeDefs: schema, resolvers })

// launch the server when the Lambda is called
exports.handler = server.createHandler();

This code will create an Apollo server and return a response, using the schema and the resolvers. You can test it in localhost:3000/dev/graphql: Apollo server test

As you can see, it works ! But our resolvers don't do anything, that's why item is null. The next step is to use real data from a database.

Get data from DynamoDB with GraphQL

I recommend using AWS DynamoDB as a database because it's very easy to setup with serverless.yml. Add the following to the existing configuration:

    ITEM_TABLE: ${self:service}-items-${self:provider.stage} # table name
    - Effect: Allow # allow the Lambda to Get and Put Items of this table
        - dynamodb:GetItem 
        - dynamodb:PutItem
      Resource: 'arn:aws:dynamodb:${opt:region, self:provider.region}:*:table/${self:provider.environment.ITEM_TABLE}'

    ItemsTable: # define the table
      Type: 'AWS::DynamoDB::Table'
          - AttributeName: itemId
            AttributeType: S
          - AttributeName: itemId
            KeyType: HASH
          ReadCapacityUnits: 1
          WriteCapacityUnits: 1
        TableName: ${self:provider.environment.ITEM_TABLE}

We are asking Serverless to create a new table for items and to give the Lambda the permissions to interact with it. More infos on the Serverless documentation.

Now we can interact with our new DynamoDB table through the aws-sdk:

yarn add -D aws-sdk

All that remains is to update our resolvers. I recommend you to declare resolvers in different files, as it'll be more maintainable. Here's our new item query:

const AWS = require('aws-sdk')

async function item(_, { id }) {
  const dynamoDb = new AWS.DynamoDB.DocumentClient()

  const params = {
    TableName: process.env.ITEM_TABLE,
    Key: {
      itemId: id,

  const { Item } = await dynamoDb.get(params).promise()

  return {
    id: Item.itemId,

Then our new createItem mutation:

const AWS = require('aws-sdk')
const { v4 } = require('uuid')

async function createItem(_, { content }) {
  const dynamoDb = new AWS.DynamoDB.DocumentClient()
  const id = v4()

  const params = {
    TableName: process.env.ITEM_TABLE,
    Item: {
      itemId: id,

  await dynamoDb.put(params).promise()

  return {

As you can see, we need to install uuid to generate an id for new items:

yarn add uuid

Your API is ready ! Let's deploy it to AWS.

Deploy an AWS Lambda with serverless framework

Deploy your new Lambda to AWS is one command away:

> serverless deploy

Serverless: Stack update finished...
endpoints: # your api endpoint !
  ANY -

Anyone can now access your API (since you configured CORS this way). Here's my version.
Test GraphQL API


Today you learnt:

  • How to build a Lambda
  • How to test it locally
  • How to build a GraphQL API on top of it
  • How to deploy it to AWS using the Serverless framework

And there's more you can do:

  • Add updateItem and deleteItem mutations
  • Add ESLint & Prettier for code readability
  • Use TypeScript instead of plain Javascript
  • Add authentication with AWS Cognito
  • etc.

If you want to learn more about Serverless, I recommend you to read the Serverless Handbook.

If you liked this article, you can follow me on Twitter as I share a lot of things related to Serverless and GraphQL.