Build a 24/7 WhatsApp Customer Service Agent with n8n and Chatmaid

Customer service has a WhatsApp problem. Your customers live on WhatsApp — but staffing a human agent to handle it around the clock is expensive, and the official WhatsApp Business API is a bureaucratic nightmare for small and mid-size teams.

The solution is an automated service agent that receives incoming messages, understands what the customer needs, and replies intelligently — running 24 hours a day without human intervention. With n8n and the Chatmaid API, you can build this in an afternoon, even if you're not a backend developer.

This guide walks you through the full setup.

What You'll Build

A WhatsApp automation workflow that:

1. Receives incoming customer messages via webhook
2. Routes them through an AI model (GPT-4o or Claude) to generate a contextual reply
3. Sends the reply back to the customer's WhatsApp number via Chatmaid
4. Logs the conversation to Google Sheets or a database

The agent runs 24/7 without supervision. When it can't confidently answer, it escalates to a human by tagging the conversation in your CRM or sending a Slack alert.

What You'll Need

• A Chatmaid Developers account (sandbox is free)
• An n8n account (n8n Cloud or self-hosted)
• An OpenAI or Anthropic API key
• A phone number connected to Chatmaid via QR scan

How Chatmaid Webhooks Work

Before building the workflow, it's important to understand what triggers it.

When a customer sends a WhatsApp message to your connected number, Chatmaid fires a message.received event to your webhook URL. The payload looks like this:

json

{ "event": "message.received", "data": { "from": "+50761234567", "to": "+15551234567", "content": "Hello, I want to know the status of my order #4892", "timestamp": "2026-06-01T14:32:00Z", "messageId": "msg_xyz789" } }

Your n8n workflow receives this, processes it, and calls the Chatmaid send endpoint to reply. That's the entire loop.

Building the n8n Workflow

Node 1: Webhook Trigger

Add a Webhook node to your workflow. Set the method to POST and copy the webhook URL — you'll paste this into your Chatmaid dashboard under webhook settings.

Important: In Chatmaid's dashboard, enable the message.received event and paste your n8n webhook URL. Chatmaid signs all webhook payloads with HMAC-SHA256 — store your webhook secret in n8n credentials and verify it in the first node if you want additional security.

Node 2: Filter — Exclude Outbound Messages

Chatmaid fires webhooks for messages you send as well as messages you receive. Add an If node that checks:

{{ $json.data.event !== "message.sent" }}

Only proceed if the event is message.received. Route the false branch to a No Operation node.

Node 3: Extract Message Content

Add a Set node to extract the fields you'll use downstream:

• customerPhone → {{ $json.data.from }}
• customerMessage → {{ $json.data.content }}
• timestamp → {{ $json.data.timestamp }}

Node 4: AI Response Generation

Add an HTTP Request node to call your AI provider. Here's the OpenAI configuration:

URL: https://api.openai.com/v1/chat/completions
Method: POST
Authentication: Header Auth with your OpenAI key

Body (JSON):

json

{ "model": "gpt-4o", "messages": [ { "role": "system", "content": "You are a helpful customer service agent for [Your Company]. Answer questions about orders, products, and policies. If you don't know the answer, say you'll connect them with a human agent. Keep responses concise and friendly. Always respond in the same language the customer writes in." }, { "role": "user", "content": "{{ $node['Extract Message'].json.customerMessage }}" } ], "max_tokens": 300 }

Extract the reply text with: {{ $json.choices[0].message.content }}

Node 5: Send WhatsApp Reply via Chatmaid

Add another HTTP Request node:

URL: https://developers-api.chatmaid.net/v1/messages/send
Method: POST
Headers:

• Authorization: Bearer sk_live_xxx
• Content-Type: application/json

Body (JSON):

json

{ "fromPhoneId": "+15551234567", "to": "{{ $node['Extract Message'].json.customerPhone }}", "content": "{{ $node['AI Response'].json.choices[0].message.content }}" }

Node 6: Log to Google Sheets (Optional)

Connect a Google Sheets node to append a row after each conversation turn:

Column

Value

Timestamp

{{ $node['Extract Message'].json.timestamp }}

Customer Phone

{{ $node['Extract Message'].json.customerPhone }}

Customer Message

{{ $node['Extract Message'].json.customerMessage }}

Agent Reply

{{ $node['AI Response'].json.choices[0].message.content }}

Adding Conversation Memory

The workflow above handles each message independently. For a more natural experience, you want the agent to remember previous turns in the conversation.

The simplest approach: store conversation history in a Google Sheet indexed by phone number. Before calling the AI, fetch the last N messages for that customer and include them in the messages array.

For production workloads, use a database. Chatmaid pairs well with Supabase — store chat history in a conversations table, query it by phone number, and pass the last 10 turns to the AI as context.

Escalation Logic

Not every message should be handled automatically. Add escalation rules:

Escalate when:

• The AI response contains phrases like "I don't know" or "I'm not sure"
• The customer has sent more than 5 messages without resolution
• The message contains trigger words like "refund", "complaint", or "urgent"

To detect these cases, add an If node after the AI Response node. Route escalations to:

• A Slack node that notifies your support team
• A Gmail node that emails the conversation log
• A Chatmaid send that tells the customer a human will follow up

Handling Media Messages

Customers often send photos, audio notes, or documents. Chatmaid's webhook payload includes a mediaUrl field when the incoming message contains media. You can pass this URL to a vision model (GPT-4o or Claude) alongside the message content:

json

{ "role": "user", "content": [ { "type": "text", "text": "{{ $json.data.content }}" }, { "type": "image_url", "image_url": { "url": "{{ $json.data.mediaUrl }}" } } ] }

This lets your agent handle photos of damaged products, screenshots of error messages, or images of receipts — without any additional infrastructure.

Testing the Workflow

1. Activate your workflow in n8n
2. Send a test message to your connected WhatsApp number from a different phone
3. Watch the execution log in n8n — you'll see each node execute in sequence
4. The reply should arrive in under 5 seconds

Use Chatmaid's sandbox environment (sk_test_*) during development. The sandbox simulates the full incoming message lifecycle, so you can trigger test webhooks without needing a real incoming message.

What This Costs

• Chatmaid: $7.99/month flat (unlimited messages)
• n8n Cloud: from $20/month (or free if self-hosted)
• OpenAI GPT-4o: ~$0.005 per 1,000 tokens — a typical customer service reply costs less than $0.001

For most small businesses, the total infrastructure cost is under $30/month. That's less than a single hour of customer service labor.