WhatsApp MCP Tool Enables AI Agent Access to WhatsApp Messages on Cloudflare

Introduction to WhatsApp MCP

WhatsApp MCP is a new server application designed to integrate WhatsApp messages with AI agents. It functions as a linked device, similar to WhatsApp Web, allowing AI clients that support the MCP protocol (e.g., Claude Code, Claude Desktop) to access and process chat data.

Functionality and Use Cases

Once linked, the tool enables AI agents to search and read WhatsApp messages, summarize conversations, and recall names associated with phone numbers. Examples include querying specific messages from contacts, finding mentions of keywords across all chats, or summarizing group activity over a period. The tool is read-only and cannot send messages.

Technical Implementation and Data Handling

The server runs within a user's Cloudflare account, connecting to WhatsApp periodically to collect new messages. These messages are then stored in a SQLite database within a Durable Object on Cloudflare. The user's phone remains the primary source of truth and continues to receive notifications independently. Data is stored exclusively within the user's Cloudflare account.

System Requirements and Setup

To deploy WhatsApp MCP, users need a Cloudflare account with the Workers Paid plan, Node.js version 20 or later, and their mobile phone. The setup process involves cloning a GitHub repository, installing dependencies, creating a KV namespace for tokens, setting an owner password, and deploying the application using Wrangler. A QR code scan or phone number option is used to link WhatsApp, and AI clients are configured to point to the deployed URL.