The classic WhatsApp MCP setup asks for a lot: Go to build the bridge, Python 3 and uv for the MCP server, FFmpeg for voice notes, and on Windows a C compiler with CGO enabled. If you just want your agent on WhatsApp, a single Node runtime is enough. Here are the options, checked against their READMEs on November 14, 2026.
Node-only servers
| Project | Install | Runtime | License |
|---|---|---|---|
| WACrew | npx -y wacrew-mcp (MCP) + collector via Docker or npm run dev | Node 20+ | Apache 2.0 |
| whatzmcp | npx -y github:whatzmcp/whatzmcp | Node 20+ | MIT |
| jlucaso1/whatsapp-mcp-ts | clone, npm install, node src/main.ts | Node 23.10+ | ISC (package.json) |
whatzmcp is the shortest path: npx -y github:whatzmcp/whatzmcp login, scan the QR code, add the same npx command to your client. It keeps everything in a local SQLite file and has sending limits built in (more in whatzmcp vs WACrew).
WACrew splits the session from the MCP: a collector that can run on a server stores messages and media, and wacrew-mcp connects to it with three environment variables. More moving parts than whatzmcp, but your agent does not miss messages while your laptop is closed, and you get a web panel, transcription and a review before sending.
whatsapp-mcp-ts is a straightforward Baileys port of the lharries tools (six of them). It uses Node's built-in TypeScript and SQLite, hence the recent Node requirement.
No local runtime at all
If you would rather not run anything on your machine:
- felipeadeildo/whatsapp-mcp deploys to your own Cloudflare account with one button and is added to clients as a remote MCP URL.
- finereli/whatsapp-mcp does the same for read-only access, and also offers a hosted beta.
Or keep Go, but hide it in Docker
whatsapp-mcp-extended runs the Go bridge and Python server in containers with docker compose. Nothing to install besides Docker. See WhatsApp MCP in Docker.
Minimal WACrew setup on one machine
git clone https://github.com/sumioshi/wacrew && cd wacrew
npm install && npm --prefix web install
cp .mcp.json.example .mcp.json
npm run dev
Open http://localhost:3000/config, scan the QR code, and Claude Code picks up the project-scoped .mcp.json. FFmpeg is required; local transcription uses MLX Whisper on Apple Silicon, or set TRANSCRIBE_BACKEND=openrouter elsewhere.
FAQ
Is Baileys as capable as whatsmeow? For reading, searching and sending text, media and voice notes, yes. Both implement the same multi-device protocol. See whatsmeow vs Baileys.
Do I still need FFmpeg? For converting audio to voice notes and extracting video frames, yes in most servers, including WACrew.
Which one has the smallest footprint? whatzmcp: one npx command and a SQLite file.



