A WhatsApp MCP server lets an AI client search your chats and send messages and media. The widely-used implementation connects to your personal WhatsApp via the multi-device web API, stores messages locally, and only surfaces them when the agent asks. Here is the setup — and the privacy caveats.
How it works
The popular whatsapp-mcp has two parts: a Go bridge that links to your WhatsApp account (QR-code login, like WhatsApp Web) and a Python MCP server that exposes tools to Claude. Messages are stored in a local SQLite DB on your machine and only sent to the LLM when a tool reads them.
Step 1 — run the bridge
Clone and start the bridge, then scan the QR code with your phone (WhatsApp → Linked Devices):
git clone https://github.com/lharries/whatsapp-mcp.git
# follow the repo: start the Go bridge, scan the QR to link your account
Step 2 — configure the client
Point your client at the Python MCP server per the repo (typically a uv/Python command):
{
"mcpServers": {
"whatsapp": {
"command": "uv",
"args": ["--directory", "/path/to/whatsapp-mcp/whatsapp-mcp-server", "run", "main.py"]
}
}
}
Restart the client; see MCP config file location.
Step 3 — verify
Ask: "Find my last conversation with Alex and summarise it." Real messages back confirm the bridge is linked.
Common problems
- QR expired / not linked — re-run the bridge and scan again; the link can drop.
- Bridge not running — the Python server needs the Go bridge alive to fetch data.
- Python/uv errors — Python 3.10+ with
uvrequired.
Privacy & security
This connects your personal WhatsApp and stores messages locally. That's powerful and sensitive: run it only on a machine you control, understand that read tools expose private chats to the LLM, and review the code before running. See how to vet an MCP server and MCP security best practices.
Going further
WhatsApp pairs with a calendar or CRM server for personal-assistant flows. Browse the communication category or curated loadouts.