Whats App Mcp Server
@meharajM
Human-in-the-loop approvals & notifications for AI agents via WhatsApp (MCP server)
Overview
What is Whats App Mcp Server?
Whats App Mcp Server is an MCP (Model Context Protocol) server that enables AI agents (like Claude or Cursor) to interact directly with users via WhatsApp. It bridges the gap between autonomous AI and the user’s phone, allowing runtime confirmations, permission requests, and status updates while away from the computer.
How to use Whats App Mcp Server?
Configure your MCP client (Cursor, Claude Desktop, etc.) to run the server via npx -y @mhrj/whatsapp-mcp and set the environment variable WHATSAPP_TARGET_NUMBER. Optionally restrict reply sources with WHATSAPP_ALLOWED_NUMBERS. First-time authentication: call the connect tool, scan the QR code with WhatsApp. Then use tools like ask_question to receive replies or send_message for one-way notifications.
Key features of Whats App Mcp Server
- Connect/disconnect to WhatsApp network
- Send one-way notifications via
send_message - Ask questions and block until a reply is received
- Concurrent questions are smartly queued and tagged
- Get connection status with
get_status
Use cases of Whats App Mcp Server
- AI agents requesting approval before deploying to production
- Remote monitoring and alerts when you are away from the IDE
- Collaborative debugging with agent asking for your input
- Fully autonomous workflows with human‑in‑the‑loop permissions
- DevOps teams integrating approval steps into agentic pipelines
FAQ from Whats App Mcp Server
How do I authenticate for the first time?
Call the connect tool. The agent returns a QR code image; scan it with WhatsApp (Linked Devices > Link a Device). The session is saved to ~/.whatsapp-mcp/baileys_auth_info.
What environment variables are required?
WHATSAPP_TARGET_NUMBER (default recipient, e.g., [email protected]). Optionally add WHATSAPP_ALLOWED_NUMBERS (comma‑separated) to restrict who can reply.
Can it be used with Cursor or Claude Desktop?
Yes. Add the server configuration to your MCP config file with npx and the required environment variables.
How do I set up autonomous “auto‑pilot” mode?
Remove IDE security blockers (always allow command/file permissions) and add a .cursorrules prompt that instructs the AI to use the ask_question tool via WhatsApp for decision points.
How are concurrent questions handled?
They are smartly queued and tagged with references, ensuring the user can reply to each in order without confusion.
