5 ways to use OAuth2 in Firebase and Google Cloud Platform

While working on QUID's core features, we're simultaneously testing them out on our own experiemental app ideas. Apps today have the luxury of being able to easily talk to each other using API's, and popular companies such as Github, Google, and Facebook make it really easy for developers to securely access a user's data. The main technology behind this is OAuth2.

This post will be looking at how to implement auth in a serverless architecture. One thing to note is I will be using client libraries and not implementing the OAuth2 mechanisms on my own. That will come in a later post.

Its important to note that whenever a third party maintained OAuth2 client is available, it should be used instead of trying to implement it on your own. OAuth2 is a standard but providers may have slight variations in their own implementations that differ from others. Instead of trying to re-invent the wheel, I prefer to just hop in the car and drive.


  • If you know nothing about OAuth2, or want a refresher, watch this video
  • Create a Firebase account.
  • Your dev environment will need Node v8 or higher and git.


The full source code can be found here. Branches were made for each section so we can following along with with git diffs. I recommended following along with the blog post/git repo, and trying to implement the flows on your own after.

The flows we will explore:

  1. Client side authorization
  2. Client side authenticated API requests
  3. Server side authenticated API requests
  4. Client side authenticated API requests using server generated access tokens
  5. Periodic API calls using Cloud Scheduler.

Let's get started!


First you will need to clone the demo repo so we can run things locally.

    git clone https://github.com/wichopy/serverless-oauth2-blog-post.git

Next, make a firebase project. Mine will be called oauth-flows.


Steps 3-5 require a billing account to be associated with your GCP project if you want to run the project on live firebase server. Otherwise, everything works inside of the cloud functions emulator.

Before deploying to live servers, add a billing account here

Install the firebase cli tool in your dev environment.

    npm install -g firebase-tools

And login

    firebase login

Note: This cli login flow is using OAuth2!

This demo was bootstrapped using firebase init to set up the initial config files, html, and cloud function tools.

In the first prompt, select functions, hosting and firestore. In the next prompt, select the project we just made, oauth-flows. The next few prompts will ask you questions based on your personal preferences. I chose the defaults for firestore rules and firestore index, javascript for cloud functions, yes to eslint, yes to installing dependencies, the default public folder for public directory, and yes to single page app.

The github repo's master branch will show what we have after this initialization. We will go into the first branch to see our first flow.

1. Client side authorization

git checkout google-signin


Run a local web server to host public/index.html and visit localhost:5000 to play with the demo.

firebase serve --only hosting

The first flow we will look at will be very simple and will set us up for the later flows. Since we are using firebase, they have a great abstraction that simplifies authorizing your users. It supports integrations with all the major players such as Facebook, Github and Google. Some powerful features include persisted auth sessions and user tables for you to manage your users.

Before being able to use one of the integrations, you must first enable it in the firebase console. Let's enable the Google auth integration by going to Authentication > Sign-in Method > Google in the Firebase console. Click the enable toggle and click save.


Let's implement a flow we are all used to seeing, the pop up sign-in.

var provider = new firebase.auth.GoogleAuthProvider();
function onSigninClick () {
  firebase.auth().signInWithPopup(provider).then(function (result) {
    console.log('Auth resposne from firebase:', result)

function onSignoutClick () {
  firebase.auth().signOut().then(function() {
    console.log('Signout successful')

Calling the signInWithPopup function will open a pop up and ask you for your Google credentials. Behind the scenes, the auth code for token exchange happens on the firebase servers which will set up a new user in your user table if one does not exist and set the auth session in your browser. Here is a diagram of what is happening.


The result from signInWithPopup contains tokens such as the access token, id token and refresh token that we'll use later.


Access Token

A short lived token that API's look for when making authenticated requests. This is only returned when a user enters their credentials (typing in the email/password or clicking on the already logged in Google account). The default expiry time for access tokens is 3600 seconds (1 hour) but we can customize this to be as long or short as our application needs.

Refresh token

A long lived token that can be use with a client secret on your backend to generate new access tokens. A default expiry for a refresh token is 6 months for Google.

ID Token

A token to identify a user in the OpenID Connect protocol, the defacto authentication framework used for single sign-on flows. This ID token can be used to authenticate users in different OAuth2 platforms, and to verify our API requests. More on this later.

Why do we have an ID token? If you watched the video linked above, its discussed that OAuth2 was designed for authentication, not authorization. The openID Connect protocol solves the authentication problem by standardizing the authentication data so all systems can talk to eachother the same way.

This is a good intro, let's do some actual API requests now.

2. Client side API requests

git checkout google-api-requests


Let's use Firebase's Google auth implmentation and add additional scopes so we can get an access token that will read from one of Google's many apis. In this example, we will read Google calendar events.

Before we can do this, we need to add the gapi javascript library to our html and enable the Calendar API in this project.
Add to our head tag:

  + <script src="https://apis.google.com/js/api.js"></script>

To add apis to a Google / Firebase project:

Adding an api to a project will let Google know that your client ID / api keys will be able to request access to the APIs we enabled.

You should see this indicator after enabling a Google api:

Here is an overview of what we'll implement.


After enabling, we will need to add the events scope to our Google auth provider.

        var provider = new firebase.auth.GoogleAuthProvider();
        const calendarEventsScope = 'https://www.googleapis.com/auth/calendar.events.readonly'
        // Add a scope for an api you want to grab data from. In this example we are reading google calendar events for your main calendar.

This scope is needed so when the user provides their credentials, they will also be notified of what data you are trying to access, which will give us permission to access their data after they click accept.

This is what the user will see when trying to login after adding the calendar events scope.



Now that we added the scope, we can add the code to make an api request after we get an access token:

// This function adds an access token to the google api client if available, otherwise it will ask you for your credentials again.
function authenticateGoogleAPI () {
 return new Promise((resolve, reject) => {
   if (!accessToken) {
     // Reentering the app as a logged in firebase user, we need to reauth to get a new access token.
     firebaseUser.reauthenticateWithPopup(provider).then(result => {
       console.log('reauthenticate result', result)
       accessToken = result.credential.accessToken
         access_token: accessToken
   } else {
     // Already have access token from logging in
       access_token: accessToken
* Google API
* */
function onGapiLoad () {
 // Enable the api you want to use in the developer console
 // https://console.developers.google.com/apis/library/
 function fetchData () {
     .then(() => {
       return gapi.client.request({
         // Pick an endpoint based on the scope and api you defined.
         path: 'https://www.googleapis.com/calendar/v3/calendars/primary/events',
         method: 'GET'
     .then(result => {
       fetchResponse.innerText = result.body
   'apiKey': 'AIzaSyBWLnDa3OxY_QNenfQ-ikkRNLur9jFsoUA',
   // clientId and scope are optional if auth is not required.
   'clientId': '315834859490-9aibgf8ofbop7h0o050nahkpb01272ac.apps.googleusercontent.com',
   'scope': 'profile https://www.googleapis.com/auth/calendar.events.readonly',
 }).then(function() {
   // 3. Initialize and add an onClick handler to make the API request.
   console.log('google api initialized')
   fetchButton.addEventListener('click', fetchData)
// 1. Load the JavaScript client library.
gapi.load('client', onGapiLoad);

The key blobs to look at are:

    access_token: accessToken


    return gapi.client.request({
      // Pick an endpoint based on the scope and api you defined.
      path: 'https://www.googleapis.com/calendar/v3/calendars/primary/events',
      method: 'GET'

setToken Add the access token to our Google client, essentially authenticating it to make api requests.

request Uses the access token that we set and adds it to request HEADers for us, abstracting this low level api so we can focus on what we really want, the data.

The request for Google calendar events was quite simple but you will need to look up the API you want to interact with to figure out what the request will look like. You may need to include required params in the body field in order to have a successful request. You should be able to find what you are looking for with a quick Google search. For example, this is what I used to construct the path seens above.

Did you notice the authenticateGoogleAPI function? We only get an access token when we enter our credentials into the pop up. If an auth session exists already when your user reenters the app, they will need to re-enter their credentials in the pop up in order to get a new access token.

This flow works for some use cases, but most likely we would want our users to just authenticate once and be able to access their data as long as they are logged in. Let's look at how to do this in the next section.

3. Server side API requests

git checkout offline-api-requests


In OAuth2 terms, being able to access a user's data while they are away from the app is called offline access. We will use this mechanism to improve our user experience.

The flow diagram below will show the grant offline access flow from when a user consents to us accessing their data to how we store a refresh token.


On the client, we will use the gapi grantOfflineAccess method to start this flow.

function openConsentWindow() {
      scope: calendarEventsScope
    }).then(res => {

Cloud Functions have entered the game

We will make our own microservice using cloud functions. These cloud functions can be run in your local env using the firebase cli command firebase serve --only functions or if you are inside of the functions folder, npm run serve. We should do all our development using the emulator so we don't eat into our quotas and if you don't have billing set up, your cloud functions cannot make api requests outside of the firebase realm.

Let's take a look at our first cloud function which will be used to accept the access code returned from the grantOfflineAccess response.

const admin = require("firebase-admin");
const { google } = require('googleapis');

// From google api credentials: https://console.cloud.google.com/apis/credentials/ , go to the web client ID and download the JSON
const googleSecrets = require("./google-secrets.json");
// From firebase console: https://console.firebase.google.com/project/[YOUR PROJECT ID]/settings/serviceaccounts/adminsdk
const serviceAccount = require("./oauth-flows-service-key.json");

const clientId = googleSecrets.web.client_id
const clientSecret = googleSecrets.web.client_secret
// Don't use an actual redirect uri from our list of valid uri's. Instead, it needs to be postmessage.
// https://stackoverflow.com/a/48121098/7621726
const redirectUri = 'postmessage'

const oauth2Client = new google.auth.OAuth2(

const app = admin.initializeApp({
  credential: admin.credential.cert(serviceAccount),
  databaseURL: "https://oauth-flows.firebaseio.com"

 // Wrap callback in a promise.
const getToken = (code) => {
  return new Promise((res, rej) => {
    oauth2Client.getToken(code, (err, token) => {
      if (err) return rej(err)

       return res(token)

exports.offlineGrant = functions.https.onRequest(async (request, response) => {
  const { code, uid } = request.query

   if (!code) {
    response.status(400).send('Missing auth code')
  if (!uid) {
    response.status(400).send('Missing uid')

   const token = await getToken(code)

   // Overwrite previous value.
  await app.firestore().collection("users").doc(uid).set({ refreshToken: token.refresh_token })
  console.log('save refresh token in user doc')


Great, we requested for a user's credentials and saved them to our database.
Now whenever we want to access their data, all we need to do is grab the refresh token, pass it to the Google api client, and call the api endpoint the token is scoped to.

Here is a diagram showing this flow.


This is what it would look like with our calendar events example.

 exports.events = functions.https.onRequest(async (request, response) => {
  const { uid } = request.query

   if (!uid) {
    response.status(400).send('Missing uid')

   const user = await app.firestore().collection("users").doc(uid).get()

   const refreshToken = user.data().refreshToken

   oauth2Client.setCredentials({ refresh_token: refreshToken })

  // https://github.com/googleapis/google-api-nodejs-client to find your api.
  // Use VS Code, the intellisense is magical.
  const calendarApiClient = google.calendar({
    version: 'v3' ,
    auth: oauth2Client,

   const events = await calendarApiClient.events.list({
    calendarId: 'primary'

   // For development only, we want to restrict this to allowed origins.
  response.set('Access-Control-Allow-Origin', '*');

Note, remember the refresh token on the firebase auth user object on the frontend? It will not work with googleapis as firebase took our ID token from the Google login and made their own custom tokens with it. We need to use a refresh token created by our auth code and client secret in order to access a user's data offline.

Authenticating Server Side API Request using ID Tokens

In the previous request, we passed a uid to our endpoint. Although this is great for getting something working quickly locally, we should not transmit a user ID this way as its not safe. How then can we authenticate a request to our server and know that it was done by someone logged into our firebase app? One way to do it is by using the ID token.

The ID Token can be verified by publicly facing cloud functions before performing an authenticated API request. There is a handy firebase method to perform this check, verifyIdToken.

const authorize = async (req, res) => {
  // https://github.com/firebase/functions-samples/blob/master/authorized-https-endpoint/functions/index.js
  console.log('Check if request is authorized with Firebase ID token');

  if ((!req.headers.authorization || !req.headers.authorization.startsWith('Bearer '))) {
    console.error('No Firebase ID token was passed as a Bearer token in the Authorization header.',
      'Make sure you authorize your request by providing the following HTTP header:',
      'Authorization: Bearer <Firebase ID Token>');

  let idToken;
  idToken = req.headers.authorization.split('Bearer ')[1];

  try {
    const decodedIdToken = await app.auth().verifyIdToken(idToken)
    console.log('ID Token correctly decoded', decodedIdToken);
    req.user = decodedIdToken;
    return decodedIdToken;
  } catch (error) {
    console.error('Error while verifying Firebase ID token:', error);

// Inside publicly accessible cloud fucntions:
const authUser = await authorize(request, response)

if (!authUser) {
  // Stop running the cloud function if unauthorized.

4. Client side authenticated API requests using server generated access tokens

git checkout request-client-access-token


Making api calls with cloud functions will add to your free quota. Depending how frequently you want to access these api calls, it might make more sense to let your frontend client make the calls using the method we outlined in Client side API Requests, by setting an access token to the gapi client. With our stored refresh token, we can now make access tokens on demand without having to ask the user to reauthenticate.

exports.tokens = functions.https.onRequest(async (request, response) => {
  const { idToken } = request.query
  response.set('Access-Control-Allow-Origin', '*');
  if (!idToken) {
    response.status(400).send('Missing id token')

   let uid
  try {
    uid = await extractUid(idToken)
  } catch (err) {
    response.status(400).send('Error verifying id token')

   console.log('uid', uid)

   const user = await app.firestore().collection("users").doc(uid).get()
  if (!user.exists) {
    response.status(400).send('No credentials saved for this user.')

  const refreshToken = user.data().refreshToken
  console.log('refresh token', refreshToken)
  oauth2Client.setCredentials({ refresh_token: refreshToken })
  const accessToken = await oauth2Client.getAccessToken()
  response.send({ accessToken: accessToken.token })

Now simply use this access token like we did in the Client Side API Requests section, except now we don't need to reauthenticate the user everytime they return to the app!

5. Periodic API Requests

git checkout scheduled-api-requests


We've covered different ways for calling API's from a frontend client and on the server, but how about if we want to call an API at a regular interval. An example of this is calling the Google fitness API everyday to get a user's previous days step count for your next awesome fitness app. With cloud functions and GCP's Cloud scheduler it couldn't be any easier.

Create PubSub Topic

In the Google Cloud console, go to Pub Sub and then the Topics section. You shouldn't see any topics here. Click on Create A Topic. We will create a topic called getEvents.


Create cloud scheduler Job

Cloud scheduler is relatively new feature in GCP that let's you create scheduled tasks for all the supported hooks. We will be making one to talk to our pub sub topic. Go to the Cloud Scheduler module and click on Create Job.

Most of the fields are self explanatory. The Frequency is written in Cron notation. To get an hourly job running, the syntax is 0 * * * *. For now we have a blank payload.


Pubsub Cloud Function

The last piece, our cloud function. The firebase cloud functions library has a few triggers we can take advantage of, with pubsub being one of them. This will get fired everytime the job runs.

exports.eventsSubscription = functions
  .onPublish(async (msg, ctx) => {
    const usersSnapshot = await app.firestore().collection("users").get()
    usersSnapshot.forEach(user => {
      const refreshToken = user.data().refreshToken
      const events = getEvents(refreshToken)
      console.log('Events for ', user.id, ': \n', events.data)

If you want to pass data in the pubsub, you can fill it in with a JSON and parse it on the receiving end using a node Buffer. It would look something like this:

In the job payload:

  "action": "NOTIFY_USERS",
  "msg": "WAKE UP!!"

In your cloud function.

onPublish((msg, ctx) => {
  let messageBody = msg.data
    ? Buffer.from(msg.data, "base64").toString()
    : null;

  messageBody = JSON.parse(messageBody);

To test the pubsub function locally, we can't use the serve command as it only serves https functions. To test, we can use the functions shell.

    cd functions # if you aren't in the functions folder already.
    npm run shell # or npm start

    # When the shell finishes booting up, event the following command:

You should see your function trigger and have it call the apis for every user you have saved in your firestore.


We live a golden age right now where its easier than ever to talk to third party APIs. With these practical examples of how to make authenticate API calls, I hope this post has inspired you with some techniques for making your next app.

In future posts I would like to expand on these topics more by:

  • Implementing our own OAuth2 mechanism using cloud functions
  • Make a medium complexity application that extensively uses OAuth2 and cloud services

Until then, happy hacking!


Author image

About William Chou

Wielder of javascript and student in the school of serverless cloud services.
  • Toronto, Ontario, Canada