Skip to content
SocialMate
Sign in Download free
MCP · Claude · Cursor

A native WhatsApp MCP server for Claude, Cursor and any agent

SocialMate ships a native Model Context Protocol server so Claude Desktop, Cursor and any MCP client can control WhatsApp as tools: send, read, contacts, groups and conversation memory. Self-hosted, no Business API.

Download SocialMate free Read the docs
In short

SocialMate speaks the open Model Context Protocol. Its native socialmate-mcp server lets Claude Desktop, Cursor, Cline — any MCP client — control WhatsApp as native tools: send and read messages, look up contacts, manage groups, queue batches of personalised messages and recall whole conversations. It runs on your own machine with your own number, and every tool call flows through the same anti-ban, scope and tier gates as the API, so an agent can never bypass a limit.

Connect WhatsApp to Claude via MCP

  1. Turn on the local API server. In the SocialMate app → API & Integrations, switch the local API server on. It is off by default, and the MCP server talks to it — so nothing works until it is running.
  2. Create an API key. Still in API & Integrations → API Keys, create a key. Its scope (read / send / admin) and your license tier decide which tools work. Copy it now — it is shown once.
  3. Add the MCP server to your client. For Claude Desktop or Cursor, add a server to claude_desktop_config.json that runs: npx -y socialmate-mcp, with SOCIALMATE_API_KEY and SOCIALMATE_BASE_URL in its env. For Claude Code, run: claude mcp add socialmate --env SOCIALMATE_API_KEY=sm_xxx --env SOCIALMATE_BASE_URL=http://127.0.0.1:3456 -- npx -y socialmate-mcp. There is nothing to install — npx fetches the package on first run.
  4. Restart the client. Reopen Claude Desktop / Cursor and the WhatsApp tools appear in the tool list. On Claude Code, run claude mcp list to confirm it connected.
  5. Talk to your WhatsApp. Ask the agent to message a contact, summarise a chat, or look someone up — it calls the SocialMate tools directly.

Connect it in two minutes

Nothing to install globally — npx fetches the server on first run. Paste your own API key from the app where the placeholder is.

Claude Desktop · Cursor · Cline — claude_desktop_config.json

{
  "mcpServers": {
    "socialmate": {
      "command": "npx",
      "args": ["-y", "socialmate-mcp"],
      "env": {
        "SOCIALMATE_API_KEY": "sm_paste_your_api_key_here",
        "SOCIALMATE_BASE_URL": "http://127.0.0.1:3456"
      }
    }
  }
}

Claude Code — one command

claude mcp add socialmate \
  --env SOCIALMATE_API_KEY=sm_paste_your_api_key_here \
  --env SOCIALMATE_BASE_URL=http://127.0.0.1:3456 \
  -- npx -y socialmate-mcp

Keep the -y: without it npx can stop to ask you to confirm the install — inside an MCP client, where there is no terminal to answer it, so the server never starts.

Full setup guide socialmate-mcp on npm

How SocialMate compares

Capability SocialMate This is us WAHA Evolution API Twilio (Cloud API)
Self-hosted, your own number Yes Yes Yes No
No WhatsApp Business API / approval Yes Yes Yes No
No per-message fees Yes Yes Yes No
Native n8n node (usableAsTool) Yes Partial Partial Yes
Native MCP server Yes Yes No No
Conversation-memory tool for agents Yes No No No
Anti-ban engine Yes No No n/a
Smart queue + scheduling Yes Partial Partial Partial

SocialMate automates the WhatsApp app on your own machine. It is not the official WhatsApp Business API and is not affiliated with WhatsApp or Meta. Provider capabilities change — verify current details with each provider.

Frequently asked questions

Is there a WhatsApp MCP server?
Yes. SocialMate ships a native Model Context Protocol server (socialmate-mcp) so Claude Desktop, Cursor, Cline and any MCP client can control WhatsApp as tools — send and read messages, look up contacts, manage groups, queue batches of personalised messages and fetch conversation memory.
How do I connect WhatsApp to Claude Desktop?
Three things, in order: switch on the local API server in the SocialMate app (API & Integrations — it is off by default), create an API key there, then add a server to your claude_desktop_config.json that runs "npx -y socialmate-mcp" with SOCIALMATE_API_KEY and SOCIALMATE_BASE_URL in its env. Restart Claude and the WhatsApp tools appear. Cursor and Cline take the same config.
How do I connect WhatsApp to Claude Code?
One command: claude mcp add socialmate --env SOCIALMATE_API_KEY=sm_xxx --env SOCIALMATE_BASE_URL=http://127.0.0.1:3456 -- npx -y socialmate-mcp. Then run "claude mcp list" to confirm it connected. Use the API key from the SocialMate app (API & Integrations → API Keys), and make sure the local API server there is switched on first — it is off by default.
Do I need to install anything?
No. The server is published to npm as socialmate-mcp and npx fetches it on first run — there is nothing to install globally. You only need Node.js 18.17 or newer, and the SocialMate app itself.
Do I need to turn anything on in SocialMate?
Yes — the local API server, and it is off by default. Open the app → API & Integrations and switch it on. The MCP server is a translator over that local API, so if it is not running your client will start fine and then fail every tool call. This is the most common reason the WhatsApp tools do not appear.
Does it use the WhatsApp Business API?
No. It drives the normal WhatsApp app on your own machine with your own number — no Meta business verification, no templates, no per-message fees.
Can the AI bypass rate limits over MCP?
No. Every MCP tool call runs through the same anti-ban engine, key scope and license tier as the HTTP API. The MCP server is only a translator — it cannot bypass a limit.
Can a Claude or Cursor agent auto-reply when a message arrives?
MCP is request/response — it has no inbound push, so an MCP client only acts when you ask it to. To react to messages automatically, either poll the whatsapp_fetch_new_messages tool on a loop, or drive the trigger from the n8n SocialMate Trigger (message.received) or a webhook to your code and let the agent act back through MCP. For a desktop assistant you drive by hand, MCP alone is perfect.
Local only, or remote too?
Both. Run it locally over stdio for Claude Desktop / Cursor on the same machine, or expose it remotely over a Pro named tunnel for a hosted agent.
Also see Building in n8n instead? See the WhatsApp AI agent tool for n8n →

Give your AI a WhatsApp

Self-hosted on your own machine and number — no WhatsApp Business API, no per-message fees. Free to start.

Download free Compare Free vs Pro
Free forever · no card Download free