whatsapp-mcp

Why we included this project

This project bridges WhatsApp to the Model Context Protocol, so agents like Claude or Cursor can actually search and act on your personal chat history. It runs a small local bridge that signs into your own WhatsApp account and stores messages in a SQLite database, then gives the agent tools to find contacts, pull up the context around a message, grab attached media, and send text or files back to individuals or groups. You could ask an assistant to dig up what a client sent last week, summarize a long group thread, or draft a reply and fire it off without leaving your desktop. Privacy is handled sensibly: messages only reach the LLM through a tool call you control, so nothing leaves the machine on its own. The honest catch is that it leans on the unofficial WhatsApp web multidevice API, so you may need to re-authenticate roughly every three weeks, and the maintainers themselves flag that a malicious project could exfiltrate your private messages. For developers who live in WhatsApp and want their conversations to be part of an agent workflow, this is a practical, self-hosted integration worth trying.