Cover image for Creating a Whatsapp chatbot using Node JS, Dialogflow and Twilio

Creating a Whatsapp chatbot using Node JS, Dialogflow and Twilio

newtonmunene_yg profile image Newton Munene Updated on ・10 min read

I'm going to walk you through creating a WhatsApp chatbot using Dialogflow. I initially wanted to do this in several blog posts but I'm going to cover it all in one. Brace yourself it might be a bit long.


  1. Twilio account
  2. Dialogflow account
  3. Node Js
  4. Javascript knowledge

Getting Started

Install the latest stable version of Node Js if you don't have it.


Visit twilio and sign up for a new account if you do not have one. You'll be given some credits to get you started. After they are over you'll need to pay to get more so use them wisely.

On your dashboard take note of your Account SID and Auth Token.
Head over to https://www.twilio.com/console/sms/whatsapp/learn and follow the instructions to connect your Whatsapp account to your sandbox. This is necessary for the sandbox environment. This is all you need for now, we'll come back to this later.


According to their website ,

Dialogflow is an end-to-end, build-once deploy-everywhere development suite for creating conversational interfaces for websites, mobile applications, popular messaging platforms, and IoT devices. You can use it to build interfaces (such as chatbots and conversational IVR) that enable natural and rich interactions between your users and your business. Dialogflow Enterprise Edition users have access to Google Cloud Support and a service level agreement (SLA) for production deployments.

We will be using Dialogflow to power our chatbot. Head over to Dialogflow Console and create a new agent. I will not dive into the specifics of creating and training agents, handling entities, intents and more. That is beyond the scope of this tutorial. You can find multiple resources on this online.

After creating your agent, click on the Small Talk tab on the left and enable it. This allows our bot to respond to small talk and common phrases. You can customize the responses on the same tab. You can do this for a more personalized experience. On the right side of your Dialogflow console, there's an input field where you can test out your bot.

When you've tested out your bot and are satisfied you can now follow the steps below to set up authentication for accessing your chatbot via an API.

  1. On your Dialogflow console, open settings by clicking on the gear icon next to our project name.
  2. Take note of the Project Id that is on the General tab of the settings page under Google Project section. We'll be using that later.
  3. Follow the link next to Service Account.

Screenshot from 2020-01-03 10-54-09.png

  • Create a new service account, give it an appropriate name.

Screenshot from 2020-01-03 11-00-06.png

  • Set Dialogflow role to Dialogflow API Admin

Screenshot from 2020-01-03 11-00-26.png

  • Create a new key and choose JSON.

Screenshot from 2020-01-03 11-00-52.png

  • Rename the downloaded JSON file to credentials.json. This is just so we can reference it easily. We will come back to this file later.

Backend (Node JS)

I will be using typescript in this project. You shouldn't feel intimidated even if you haven't used it before. You can check out this guide on how to get started with typescript and express.

Open your terminal and create a new project.

mkdir wa-chatbot && cd wa-chatbot

Initialize a new node js project inside the folder we just created and changed directory into

npm init -y

Install the following dev dependencies

  1. nodemon
  2. typescript
  3. ts-node
  4. ts-lint
npm i -D nodemon typescript ts-node ts-lint

Install the following dependencies

  1. express
  2. dotenv
  3. twilio
  4. dialogflow
  5. @overnightjs/core
  6. @overnightjs/logger
  7. body-parser
  8. cors

We're using Overnight to stay closer to the MVC pattern and utilize Object Oriented style of programming. Read more about Overnight here .

npm i -S express dotenv twilio dialogflow @overnightjs/core @overnightjs/logger body-parser cors

We also need to install types for these modules

npm i -D @types/node @types/express @types/twilio @types/dialogflow @types/body-parser @types/cors

Next, we'll create a tsconfig.json file. In the root of your project make a new file.

touch tsconfig.json

Copy and paste the following content inside the new file

  "compilerOptions": {
    "module": "commonjs",
    "strict": true,
    "baseUrl": "./",
    "outDir": "build",
    "removeComments": true,
    "experimentalDecorators": true,
    "target": "es6",
    "emitDecoratorMetadata": true,
    "moduleResolution": "node",
    "importHelpers": true,
    "types": ["node"],
    "typeRoots": ["node_modules/@types"]
  "include": ["./src/**/*.ts"],
  "exclude": ["./src/public/"]

I will not go into the specifics for now but you can read typescript documentation for more information.
Next, create a tslint.json file at the root of your project and paste the following content inside.

  "defaultSeverity": "warning",
  "extends": ["tslint:recommended"],
  "jsRules": {},
  "rules": {
    "trailing-comma": [false],
    "no-bitwise": false,
    "jsdoc-format": true,
    "deprecation": true,
    "interface-name": true,
    "no-duplicate-imports": true,
    "no-redundant-jsdoc": true,
    "no-use-before-declare": true,
    "variable-name": false,
    "object-literal-sort-keys": false,
    "member-ordering": true,
    "await-promise": true,
    "curly": true,
    "no-async-without-await": true,
    "no-duplicate-variable": true,
    "no-invalid-template-strings": true,
    "no-misused-new": true,
    "no-invalid-this": true,
    "prefer-const": true
  "rulesDirectory": []

Let's set up our backend structure.
Open your terminal and run the following commands inside the root of your project.

mkdir src && touch src/AppServer.ts && touch src/start.ts

AppServer.ts is where we will set up our express app.
Paste the following inside src/AppServer.ts

import * as bodyParser from "body-parser";
import * as controllers from "./controllers";
import { Server } from "@overnightjs/core";
import { Logger } from "@overnightjs/logger";
import * as cors from "cors";
export class AppServer extends Server {
  private readonly SERVER_STARTED = "Server started on port: ";

  constructor() {
    this.app.use(bodyParser.urlencoded({ extended: true }));

  private setupControllers(): void {
    const ctlrInstances = [];
    for (const name in controllers) {
      if (controllers.hasOwnProperty(name)) {
        const controller = (controllers as any)[name];
        ctlrInstances.push(new controller());

  public start(port: number): void {
    this.app.get("*", (req, res) => {
      res.send(this.SERVER_STARTED + port);
    this.app.listen(port, () => {
      Logger.Imp(this.SERVER_STARTED + port);

In this file, we set up our AppServer class which extends the Server class from Overnight. In the constructor we initialise the Server class passing true as a parameter. We then head on to configure some middlewares for our app. We might be receiving JSON data in our requests so we use body-parser to ensure it's handled properly.

We then set up our controllers which we will create in a short while. After this, we define the start method which will start up the app on the port passed to it as a parameter

Paste this into src/start.ts. This is the starting point for our application.

import { config } from "dotenv";
import { AppServer } from "./AppServer";

const appServer = new AppServer();

At the top, we import config from dotenv and call it. We use this to configure our environment variables and load them into process.env. We also initiate a new instance of the server and call the start method passing in a port to use.

At this point, we need to set up our controllers. Create a folder inside src and call it controllers. Inside src/controllers create two files: BotController.ts and index.ts

Inside BotController.ts paste the following code

import { Request, Response } from "express";
import { Controller, Post } from "@overnightjs/core";
import { Logger } from "@overnightjs/logger";

export class BotController {


You'll notice some weird syntax just before our controller class. That's a decorator. We use it to tell the compiler that our class is a controller. We also pass an argument which is our URL path. With this, we can now make restful requests to [SERVER]:[PORT]/api/bot.

We don't have any routes defined yet. For Twilio, we will only need a POST route. Inside the BotController class add the following code.

  private postMessage(request: Request, response: Response) {
    Logger.Info("A post request has been received");
    return response.status(200).json({
      message: "A post request has been received"

You'll notice another decorator which tells our compiler that the method handles POST requests.

In src/controllers/index.ts add the following code. This exports our controllers so that it will be easy to export any future controllers.

export * from "./BotController";

The fun stuff

It's time to get to the fun stuff. Let's set up our app to communicate with Twilio and Dialogflow.
Create a folder called utils under src. Inside utils create two files: dialogflow.ts and twilio.ts

Inside Dialogflow.ts:

// dialogflow.ts

const dialogflow = require("dialogflow");
const credentials = require("../../credentials.json");

const sessionClient = new dialogflow.SessionsClient({
  credentials: credentials
const projectId: string = process.env.DIALOGFLOW_PROJECT_ID!;

export const runQuery = (query: string, number: string) => {
  return new Promise(async (resolve, reject) => {
    try {
      // A unique identifier for the given session
      //const sessionId = uuid.v4();
      const sessionId = number;
      // Create a new session

      const sessionPath = sessionClient.sessionPath(projectId, sessionId);

      // The text query request.
      const request = {
        session: sessionPath,
        queryInput: {
          text: {
            // The query to send to the dialogflow agent
            text: query,
            // The language used by the client (en-US)
            languageCode: "en-US"

      // Send request and log result
      const responses = await sessionClient.detectIntent(request);

      const result = responses[0].queryResult;

    } catch (error) {

Here we're importing Dialogflow and also the credentials.json file we downloaded when setting up our chatbot on Dialogflow. Remember that file? Move it to your project's root folder. We're the setting up a new SessionsClient using the credentials file. In our runQuery function we're taking in a query to send to Dialogflow and also the user's Whatsapp number which we will use to set up a Dialogflow session unique to that user. We then send the query to Dialogflow and return the response.

In twilio.ts add the following code :

import { Twilio } from "twilio";
const accountSid = process.env.TWILIO_ACCOUNT_SID!;
const authToken = process.env.TWILIO_AUTH_TOKEN!;

const client = new Twilio(accountSid, authToken);

export const sendMessage = (to: string, from: string, body: string) => {
  return new Promise((resolve, reject) => {
      .then(message => {
      .catch(error => {

Here we create a new Twilio client and instantiate using our Twilio Account SID and Auth Token. We then call the client.messages.create function which takes in the number of the user, the number sending the message(in this case, the Twilio sandbox number) and also a body. We then return the message Id.

You've probably noticed we've used a few environment variables that we haven't defined yet. In the root of your project create a .env file. Inside paste the following code and make sure to replace placeholders with appropriate values. I asked you to take note of the values required at some point in this tutorial.


Go back to BotController.ts and replace the postMessage method with the following code.

  private postMessage(request: Request, response: Response) {
    // Here we get the message body, the number to which we're sending the message, and the number sending the message.
    const { Body, To, From } = request.body;

    // Here we're sending the received message to Dialogflow so that it can be identified against an Intent.
    runQuery(Body, From)
      .then((result: any) => {
        // We send the fulfilment text received back to our user via Twilio
        sendMessage(From, To, result.fulfillmentText)
          .then(res => {
          .catch(error => {
            console.error("error is ", error);
      .catch(error => {
        console.error("error is ", error);
    return response.status(200).send("SUCCESS");

Twilio hits this method when it receives a message from Whatsapp. We extract the message body, the sender, and the recipient(in this case, Twilio sandbox number). We then send the received body to Dialogflow and get a fulfilment text. We use the Twilio client we set up earlier to send back the fulfilment text to the user.

Now there's only one more thing left to do. Open up your package.json and replace the scripts with the following

"scripts": {
    "tsc": "tsc",
    "prestart": "npm run build",
    "dev": "ts-node src/start.ts",
    "dev:watch": "nodemon",
    "build": "rm -rf ./build/ && tsc",
    "start": "node build/start.js"

The full file looks this

  "name": "wa-chatbot",
  "version": "1.0.0",
  "description": "",
  "main": "build/start",
  "scripts": {
    "tsc": "tsc",
    "prestart": "npm run build",
    "dev": "ts-node src/start.ts",
    "dev:watch": "nodemon",
    "build": "rm -rf ./build/ && tsc",
    "start": "node build/start.js"
  "keywords": [],
  "author": "",
  "license": "ISC",
  "devDependencies": {
    "@types/body-parser": "^1.17.1",
    "@types/cors": "^2.8.6",
    "@types/dialogflow": "^4.0.4",
    "@types/express": "^4.17.2",
    "@types/node": "^13.1.2",
    "@types/twilio": "^2.11.0",
    "nodemon": "^2.0.2",
    "ts-lint": "^4.5.1",
    "ts-node": "^8.5.4",
    "typescript": "^3.7.4"
  "dependencies": {
    "@overnightjs/core": "^1.6.11",
    "@overnightjs/logger": "^1.1.9",
    "body-parser": "^1.19.0",
    "cors": "^2.8.5",
    "dialogflow": "^1.0.0",
    "dotenv": "^8.2.0",
    "express": "^4.17.1",
    "twilio": "^3.39.1"

You can run npm install again in case I missed any dependencies. We also need to set up nodemon. Create a nodemon.json in your project's root folder and paste the following inside.

  "watch": ["src"],
  "ext": "ts",
  "ignore": ["src/public"],
  "exec": "NODE_ENV=development ts-node src/start.ts"

You can now run npm run dev:watch and see if your project runs successfully. Next, we need to expose our local server. Ngrok is one of the best open-source solutions but you can use whatever you prefer.

Copy your exposed server URL and open Twilio Whatsapp Sandbox. Replace the URL in WHEN A MESSAGE COMES IN with your exposed URL. Don't forget to add the path to our bot controller. i.e. /api/bot

Screenshot from 2020-01-03 11-42-28.png
Save the new changes and you can now send a message to the Twilio Sandbox Number and you'll see the response.


Here's some homework for you. Create a new controller to handle POST requests and pass the URL to the STATUS CALLBACK URL input on the Twilio Whatsapp Sandbox. Be creative and be notified when a user reads your message and when it's delivered.

The full source code can be found here https://github.com/newtonmunene99/wa-chatbot


markdown guide

Hi Newton, great article!
How can I identify the name of the itent that matched the dialogflow and give a specific answer?

I know the value in the dialogflow is at
body.intent.displayName, but how can I capture this value in the code to give a dynamic response?

Thank you!


Hey sorry for the late reply. I'm not sure I understand what you mean, but I'll answer what I understood.
When your fulfilment webhook is called the intent name will be in body.queryResult.intent.displayName you can use a switch case statement to check the name of the intent and perform some actions. You then have to return a webhook response for the request.

Here's how you can do this in express

app.post('/webhook', (req, res) => {
  switch (req.body.queryResult.intent.displayName) {
    case 'INTENTNAME':{
        const customParameter = req.body.queryResult.parameters.mYCustomParameter;
        fulfillmentMessages: [
            text: {
              text: [`My custom response based on ${customParameter}`],


Please read more about this here cloud.google.com/dialogflow/docs/f...


Hello @newton , thank you very much for your answer!
I'm having trouble understanding which part of your project I should create this code that you informed.
I understand how the webhook should work, but I don't know where I put that part of the code that you passed.

I must create a new route in the bot.js file
router.post (/webhook, (request, response) => {}); ?

and then refer to BotController.ts?
With the code you passed?

Sorry for the long text, if you have any code examples that are working, you can share that I analyze and try to understand, thank you my friend!

From my understanding, what you're looking to achieve is something like this:

  1. User send message from whatsapp
  2. You get the message and send it over to dialogflow for intent matching and await a response
  3. Use dialogflow fulfilment webhook to do something in your server
  4. Send back a response for the fulfilment webhook
  5. Receive the response from Dialogflow (this is the response you're awaiting in step 2)
  6. Send the response back to whatsapp.

I have updated the repo to include a controller for Dialogflow webhook. github.com/newtonmunene99/wa-chatbot

When you receive the webhook from dialogflow, do whatever you need to do with the intent and parameters and then return the response to dialogflow. In BotController.ts, or wherever you execute the runQuery function, just await the result then send that to Whatsapp.

Thank you so much @newton !
I'll look at it right now, have a great day my friend!


Hello, this is more or less what I need to do,

I have a web service that receives four input parameters and returns an object on the output, I need to return the value of that object to WhatsApp
My doubt is in which part of your project I put the code you passed, so when the dialogflow returns the parameters I make the call to the Webservice and use its output on WhatsApp
Thank you!


Hi Newton. Thanks for the tutorial. I seem to be having problems with the code, when i try to run the main file, it says 502 - Bad Gateway and the app won't even run.


Let me look at this. Thanks


You're welcome. Thanks, also awaiting your response on this.

Hi Tinashe, have you tried the sample code? Here's the repository github.com/newtonmunene99/wa-chatbot

Hi Newton. Yes, I cloned the code from github and only changed the Credentials but it's not working.

Did you rename .env.copy to .env.
Also what OS are you running?

yes, I renamed it. I'm running Windows 10

Please tell me if it runs with npm run dev:watch or npm run dev

The issue might be that rm isn't supported by Windows OS

Thanks mate. I was able to debug it and now it's running. Several factors were taking play, one of which is the issue of platform compatibility. Thanks so much for the assistance

Hi Tinashe, how did you solve your issue, I'm also getting the same 502 Bad Gateway error

Hi Chris. First, check if you have all of the required node modules installed. The second check is this: if you're testing your project in a local environment, you must run npm run dev first, then once the server is running, you also start any other service or app that wants to use your project.


Hello @newtonmunene_yg , how are you my friend?
Do you know if it is possible to send an image as a reply on WhatsApp?
Currently I receive the parameters through Dialogflow, process in the webhook and respond with some text back to WhatsApp, I would like to know if it is possible to respond with an image stored locally for example.

I saw it here, but I couldn't make it work

Thank you!


Thanks to you I could understand the logic and create my own project.


This looks like a great article. I've bookmarked it to come back to when I've got some time!


Hi Newton, thanks for your help, I have a question when you receive a message and you return the answer successfully, twilio doesn't mark you as a mistake?


Hi Newton, Can you please detail more how to expose the local server with Ngrok. I'm totally new in this area, thank you.


Hello Nicola. You need to first install ngrok on your machine. Check out ngrok.com/ for instructions on this. Once you have this installed, run the Node Js project using npm run dev. Take note of the port that it's running on, I believe its 3000 for this particular project. Next you need to expose your local webserver. Read about that here, dashboard.ngrok.com/get-started/tu... . The command you will need to run will be something like ./ngrok http 3000. 3000 here being the port the project is running on. Ngrok will expose your local server and give you a couple of links that are accessible on the internet. Copy the https link and thats what you will paste in the Twilio Whatsapp Sandbox page (twilio.com/console/sms/whatsapp/sa... ). Don't forget to add the bot's route. ie /api/bot


I'm sorry Newton for bothering you, When I run "npm run dev" it looks like I miss the file package.json (See log below)

In your article you talk about a package.json file where to add scripts. Are you referring to the package.json I find in the dialoflow fullfilment section or is it another file?

I think I'm missing something very important here. Sorry...

npm ERR! path /home/nicolasoccal/wa-chatbot/package.json
npm ERR! code ENOENT
npm ERR! errno -2
npm ERR! syscall open
npm ERR! enoent ENOENT: no such file or directory, open '/home/nicolasoccal/wa-chatbot/package.json'
npm ERR! enoent This is related to npm not being able to find a file.
npm ERR! enoent
npm ERR! A complete log of this run can be found in:
npm ERR! /home/nicolasoccal/.npm/_logs/2020-04-20T18_53_19_411Z-debug.log


Hi Newton. Thank you. Can you detail how to expose the local server with Ngrok?
I developed quite powerful agent in dialogflow but here I'm totally new. Thanks a lot if you can help.

Sloan, the sloth mascot Comment marked as low quality/non-constructive by the community View code of conduct

Amazing article! Ever wondered where do these different bots come from? One is a FAQ and the other one is talking to you! Who are making these bots and why do different businesses use different ones? Well, all the answers are here! In the below blog, we’re going to talk about what bot templates are, different types of templates, why are they used and what is bot marketplace! Try out the bot marketplace now s.engati.com/1p5