Skip to setup
LangGraph × WhatsApp
What’s includedHow it worksExtensionsFAQSetup
GitHub

Local setup

Link your agent’s WhatsApp.

Scan a QR code with the agent’s phone, send a test message from another account, then switch on your agent. This page runs on your computer, and API keys stay on the server.

Before you start

  • 1. Install and create your settings. Once, from the project folder. This makes a private .env file with the secrets Docker needs.
    $npm ci && npm run setup
  • 2. Start the backend. Docker runs the WhatsApp gateway, the adapter and your LangGraph agent.
    $docker compose up -d --build
  • 3. Have two WhatsApp accounts ready. One becomes the agent (a spare is best). The other sends it test messages, because the agent ignores its own.
  • No API keys needed to start. Replies are scripted until you add your keys in step 3 below.

Run it locally

Pairing happens on your computer.

This website introduces the project. To link WhatsApp, run the starter locally and open its setup page. Live mode needs OpenAI and Cloudflare keys. Fixture mode makes no provider calls.

$npm ci$npm run setup$npm run web:install$docker compose up -d --build$npm run web:dev

You’ll need Node.js 24+ and Docker Compose.

Open local setup

Under the hood

What this page connects.

LangGraph runs your agent. The starter carries WhatsApp messages, voice transcripts and photos to it, keeps each conversation separate, and sends replies to the right person.

Follow one messageWhat’s included

Stuck on a step?

I can help you get WhatsApp linked and your agent replying.

Schedule a 15-min chat or email ahmadshaukat_4@outlook.com
How it worksSetupFAQGitHub
Built with LangGraph, LangChain and WAHA by Ahmad.