Installing MCP for WhatsApp

This guide will help you connect our service’s MCP to an AI agent and start sending WhatsApp messages directly—from the AI chat, via an IDE, or through automation systems.
You can choose the most convenient connection method depending on your use case, from a quick start to advanced automation.

‍

What You Get After Connecting

After installing MCP, the AI agent will be able to:

• send WhatsApp messages to a phone number;
• use templates or generate text dynamically;
• receive delivery status;
• work in conjunction with CRMs, websites, and internal systems;
• be part of automated workflows.

‍

Option 1. Connecting to an AI Agent (Recommended)

This option is suitable if you already use an AI assistant (Claude, ChatGPT, Cursor, etc.) and want to add WhatsApp as a tool.

Connection Steps

1. Open the integration settings of your AI agent.
2. Add a new MCP server.
3. Specify the MCP URL of our service:
   Endpoint: https://mcp.chatarchitect.com/mcp
4. To pass app:secret, add an HTTP header named WabaAppSecret with the value format <app>:<secret>.
   Example:
   WabaAppSecret: nash_app:nash_secret
5. Complete the setup.

After that, WhatsApp will be available as a tool for the AI agent.

‍

Option 2. Connecting via Automation Systems (n8n)

This option is suitable if you use n8n or similar platforms (Make, Zapier, internal workflow engines).

How It Works

• The AI agent forms an intent to send a message.
• n8n manages business logic and conditions.
• Our service’s MCP executes the WhatsApp message delivery.

Quick Start with n8n

1. Obtain your app:secret in the service’s admin panel.
2. Import the starter workflow into n8n (Cloud or self-hosted).
3. Set the MCP endpoint to https://mcp.chatarchitect.com/mcp and add an HTTP header named WabaAppSecret with the value <app>:<secret>.
4. Run the workflow or adapt it to your processes.

Suitable for:

• complex action chains;
• logging and auditing;
• scaling AI-driven scenarios.

‍

Option 3. Connecting in an IDE (Cursor, Windsurf)

For developers, MCP can be connected directly inside an AI-enabled IDE.

Configuration Example

{
  "mcpServers": {
    "whatsapp": {
      "url": "https://mcp.chatarchitect.com/mcp"
    }
  }
}

‍

After saving the settings, restart the IDE and complete authorization.

‍

Option 4. Connecting via ChatGPT (Developer Mode)

1. Open ChatGPT.
2. Enable Developer Mode in settings.
3. Create a new MCP connector.
4. Specify the MCP endpoint https://mcp.chatarchitect.com/mcp and add an HTTP header named WabaAppSecret with the value <app>:<secret>.
5. Complete authorization.

Now the AI can send WhatsApp messages directly from the chat.

‍

Verification

After connecting, try the following commands:

• “Send a WhatsApp message to a client with number +7…”
• “Remind the client about the meeting tomorrow at 12:00.”
• “Notify the user that the request has been accepted.”

If the message is sent successfully, MCP is installed correctly.

‍

Security and Control

• all AI actions are logged;
• message types can be restricted;
• confirmations and rate limits are supported;
• MCP can be disabled at any time.

‍

You’re Ready

Your AI agent now not only analyzes and recommends but also acts—sending WhatsApp messages automatically and intentionally.
If you need help with setup or custom scenarios, the ChatArchitect team is always ready to assist.

‍