Creating a WhatsApp AI Assistant Using n8n: A Step-by-Step Guide

Build your own AI-powered WhatsApp chatbot using n8n, WhatsApp Business Cloud API, and OpenAI. This guide walks you through every step—from setup to testing—with real-world error handling, solutions, and an example production-ready workflow.

1. Introduction

Want to chat with an AI on WhatsApp? In this tutorial, you’ll learn how to build a WhatsApp AI Assistant using:

n8n (automation tool)
WhatsApp Business Cloud API
OpenAI (for generating intelligent replies)

By the end, you’ll have a working chatbot and gain hands-on experience with APIs, webhooks, and automation.

2. Prerequisites

n8n account (Cloud or self-hosted)
Meta Developer account with WhatsApp Business Cloud API access
OpenAI API key
Basic understanding of APIs and webhook workflows

3. Registering Your WhatsApp Business App

A. Create WhatsApp App in Meta Developer

Visit Meta for Developers
Create a new app: choose Business → WhatsApp
Link or create a WhatsApp Business Account

B. Obtain Testing Credentials

Your app dashboard will show:

Test phone number
Phone Number ID
Temporary access token (valid for only 24 hours)

Tip: For long-term use, generate a 60-day system-user token later.

C. Add Recipients to Test List

By default, only approved numbers can receive messages:

Navigate to WhatsApp → API Setup
Add numbers in E.164 format (e.g., +923001234567)
Users must accept the invite via WhatsApp to become valid recipients

4. Configuring Your Webhook in n8n

A. Create a Webhook Node

Method: GET (for initial verification)

Endpoint example:

https://yourname.app.n8n.cloud/webhook/your-unique-id/webhook

B. Verify the Webhook with Meta

In your app’s Webhook section:

Callback URL: your n8n webhook URL
Verify Token: any secret string you choose (e.g., mySecret2025)

C. Echo Back Meta’s Challenge

Configure n8n’s Webhook node response:

Field	Value
Response Mode	On Received
Response Body	`{{$json["query"]["hub.challenge"]}}`

This ensures Meta can verify your endpoint successfully.

5. Processing Incoming Messages

WhatsApp sends JSON data with structure like:

{
  "entry": [
    {
      "changes": [
        {
          "value": {
            "messages": [
              {
                "from": "923001234567",
                "text": { "body": "Hello bot!" }
              }
            ],
            "metadata": {
              "phone_number_id": "698352170035199"
            }
          }
        }
      ]
    }
  ]
}

Extract:

from: user’s number
text.body: user’s text
metadata.phone_number_id: correct sender ID

6. Integrating OpenAI for Responses

Obtaining Your OpenAI API Key

Before integrating OpenAI into your n8n workflow, you’ll need to get an API key from OpenAI.

OpenAI Dashboard Screenshot

Step-by-Step:

Sign up or log in to OpenAI.
Navigate to your API Keys page.
Click “Create new secret key”.
Optionally name your key, then copy it immediately (you won’t be able to view it again).
In n8n:
- Go to Credentials → Add New → choose OpenAI or HTTP Request.
- Paste your API key into the key field.
- Save the credentials.

Security Tip: Keep your key private. Do not share it or commit it to public repositories.

Use an HTTP Request node to call OpenAI:

POST https://api.openai.com/v1/chat/completions
Authorization: Bearer YOUR_OPENAI_API_KEY
Content-Type: application/json

{
  "model": "gpt-4o-mini",
  "messages": [
    { "role": "system", "content": "You are a helpful WhatsApp AI assistant." },
    { "role": "user", "content": "{{ $json[...] }}" }
  ]
}

Replace {{ $json[...] }} with the actual path to the user’s message text from the Webhook node.

7. Sending Replies via WhatsApp

Use another HTTP Request node to respond:

POST https://graph.facebook.com/v21.0/{{ $json[...] }}/messages
Authorization: Bearer YOUR_LONG_LIVED_TOKEN
Content-Type: application/json

{
  "messaging_product": "whatsapp",
  "to": "{{ $json[...] }}",
  "text": {
    "body": "{{ $node['OpenAI Response'].json.choices[0].message.content }}"
  }
}

Use the metadata’s phone_number_id for the endpoint and from for the recipient. This avoids hardcoding and ensures proper routing.

8. Example n8n Workflow

Here’s the actual n8n workflow I used for my WhatsApp AI assistant. It integrates a product brochure PDF into a vector store for AI-powered Q&A, and handles WhatsApp message flow.

n8n WhatsApp AI Assistant Workflow

Step-by-Step Breakdown:

1. Download Product Brochure PDF

Downloads the brochure from a given URL.
Extracts text from the PDF.

2. Create Product Brochure Vector Store

Splits large text into chunks.
Generates embeddings using OpenAI.
Saves them in a vector store for fast semantic search.

3. Use the WhatsApp Trigger

Listens for incoming WhatsApp messages.
Routes them to supported or unsupported handlers.

3a. Handle Unsupported Messages

Replies with a friendly error if the message type is not text.

4. Sales AI Agent Responds

Uses OpenAI with memory + vector store retrieval to answer based on brochure content.

5. Reply to WhatsApp User

Sends the AI-generated message back to the sender.

Why this is effective:

Context-aware answers via buffer memory.
Reduced hallucinations thanks to vector store grounding.
Smooth error handling for unsupported message types.

9. Common Errors & Fixes

Recipient phone number not in allowed list
→ Add as test number or switch to Live mode

401 – Session expired
→ Refresh token via Graph API:

GET https://graph.facebook.com/v21.0/oauth/access_token
?grant_type=fb_exchange_token
&client_id=YOUR_APP_ID
&client_secret=YOUR_APP_SECRET
&fb_exchange_token=YOUR_CURRENT_TOKEN

Webhook verification failed
→ Ensure verify token matches between Meta and n8n and echo hub.challenge
No execution data available
→ Trigger workflow via actual WhatsApp message, not manual run

10. Going Live

Add a Privacy Policy URL in Meta App → Settings → Basic (required for live access)
Switch app to Live mode once all compliance items are met
Remove restricted recipient list
Use WhatsApp message templates for messages sent after 24 hours of user interaction

11. Conclusion & Next Steps

Congrats! You now have a WhatsApp AI Assistant built with n8n and OpenAI.

Where to go from here:

Wire up custom knowledge (PDFs, documents)
Implement memory for conversation context
Launch multilingual support
Export n8n workflow as JSON for reuse

Need help? Join the n8n Community Forum or OpenAI Discord to connect with fellow builders.

Happy automating!

Creating a WhatsApp AI Assistant Using n8n: A Step-by-Step Guide#

1. Introduction#

2. Prerequisites#

3. Registering Your WhatsApp Business App#

A. Create WhatsApp App in Meta Developer#

B. Obtain Testing Credentials#

C. Add Recipients to Test List#

4. Configuring Your Webhook in n8n#

A. Create a Webhook Node#

B. Verify the Webhook with Meta#

C. Echo Back Meta’s Challenge#

5. Processing Incoming Messages#

6. Integrating OpenAI for Responses#

Obtaining Your OpenAI API Key#

Step-by-Step:#

7. Sending Replies via WhatsApp#

8. Example n8n Workflow#

Step-by-Step Breakdown:#

1. Download Product Brochure PDF#

2. Create Product Brochure Vector Store#

3. Use the WhatsApp Trigger#

3a. Handle Unsupported Messages#

4. Sales AI Agent Responds#

5. Reply to WhatsApp User#

9. Common Errors & Fixes#

10. Going Live#

11. Conclusion & Next Steps#

Where to go from here:#