Build a Grok bot for WhatsApp

What you need

• A Flow API number linked to WhatsApp (7-day free trial)
• An xAI API key from console.x.ai
• Either n8n, or a small server that can receive a webhook

Fastest: built-in AI auto-reply

No n8n, no server. In the Flow API console open your number → AI auto-reply, choose Grok, paste your xAI API key, write your instructions and turn it on. Every incoming message wakes the bot: it reads the recent chat, shows "typing…" and replies. When you answer a chat yourself, the bot stays quiet there for 30 minutes.

Option A: no code, with n8n

1. 1Add a Flow API Trigger (community node n8n-nodes-flowapi) set to Incoming Message. Activating the workflow sets your webhook automatically.
2. 2Add an AI Agent node with the xAI Grok Chat Model and your xAI API key.
3. 3In the agent's system message, describe your business, opening hours and tone, and tell it when to hand over to a person.
4. 4Add a Flow API node: Send Text Message to {{ $json.chatId }} with the agent's output.
5. 5Activate, then message your number from another phone.

Option B: 30 lines of Python

Set your number's Webhook URL in the Flow API console to this server's /whatsapp address, then run:

import os, requests
from flask import Flask, request

app = Flask(__name__)
SEND = "https://flowapi.pro/waInstance{idInstance}/sendMessage/{apiTokenInstance}"
PROMPT = "You are the WhatsApp assistant of Bright Dental Clinic. Answer briefly and kindly."

@app.post("/whatsapp")
def whatsapp():
    n = request.get_json()
    if n.get("typeWebhook") != "incomingMessageReceived":
        return "ok"
    chat = n["senderData"]["chatId"]
    text = n["messageData"].get("textMessageData", {}).get("textMessage", "")
    if not text or chat.endswith("@g.us"):
        return "ok"
    r = requests.post("https://api.x.ai/v1/responses",
        headers={"Authorization": f"Bearer {os.environ['XAI_API_KEY']}"},
        json={"model": "grok-4.3", "input": [
            {"role": "system", "content": PROMPT},
            {"role": "user", "content": text}]}, timeout=60).json()
    reply = next(c["text"] for o in r["output"] if o["type"] == "message"
                 for c in o["content"] if c["type"] == "output_text")
    requests.post(SEND, json={"chatId": chat, "message": reply}, timeout=30)
    return "ok"

Make it feel human

• Call sendTyping before the reply so the customer sees "typing…".
• Keep answers to a few short lines; WhatsApp is a chat.
• Give the bot a clear hand-off: "A team member will reply shortly" when it is unsure.

Let Grok act, not just answer

Instead of a fixed reply, give Grok the Flow API MCP link as a remote MCP tool. It can then read the conversation history, send your price list as a PDF or create a group with your team, on its own.

Questions