Vapi AI Phone Agent: Never Miss a Call at Your Salon or Coffee Shop

What Vapi Is — And Why It Is Different From a Simple Chatbot

• Speech-to-text (transcription of what the caller says in real time)
• LLM reasoning (Claude, GPT-4o, or Llama 3 — your choice of model)
• Text-to-speech (voices from ElevenLabs, Deepgram, or Cartesia)

Vapi Pricing in 2026

• Free tier: Up to 10 minutes per month for testing and prototyping
• Pay-as-you-go: approximately $0.05 per minute (voice and LLM costs combined)
• Phone numbers: $2 per month per number via Vapi, or forward your existing number for free
• Webhooks and API access: Included at all tiers

Step-by-Step: Setting Up Your Vapi Phone Agent

Step 1: Create a Vapi Account and Get Your API Key

export VAPI_API_KEY="YOUR_VAPI_API_KEY"

# Test your key with a simple API call
curl https://api.vapi.ai/assistant \
  -H "Authorization: Bearer $VAPI_API_KEY" \
  -H "Content-Type: application/json"

Step 2: Create an Assistant via the Vapi Dashboard

{
  "name": "Bella Hair Studio Receptionist",
  "model": {
    "provider": "openai",
    "model": "gpt-4o-mini",
    "temperature": 0.6,
    "systemPrompt": "You are the friendly phone receptionist for Bella Hair Studio, a premium hair salon in Austin, Texas. Your job is to answer questions about our services and hours, help callers book appointments, and provide a warm, professional experience.\n\nSalon details:\n- Address: 1420 South Congress Ave, Austin TX 78704\n- Hours: Mon, Wed–Sat 9am–7pm, Sunday 10am–5pm. Closed Tuesday.\n- Phone: (512) 555-0192\n- Services and prices:\n  * Haircut (women) — from $65\n  * Haircut (men) — from $45\n  * Balayage — from $180\n  * Full color — from $120\n  * Highlights — from $95\n  * Keratin treatment — from $250\n  * Blowout — $55\n- Stylists: Maria (color specialist), Jordan (cuts and styling), Priya (extensions and keratin)\n- Booking policy: 50% deposit required for color services, 24-hour cancellation notice required\n\nYour rules:\n- Always confirm the caller's name and best phone number before finalizing any booking\n- If a caller asks something you cannot answer confidently, offer to have someone call them back within the hour\n- Never make up availability — always use the check_availability tool before quoting open slots\n- Always confirm at the end of a booking that the caller will receive a confirmation text\n- If a caller sounds upset or frustrated, offer to transfer them to the owner immediately"
  },
  "voice": {
    "provider": "elevenlabs",
    "voiceId": "rachel",
    "stability": 0.5,
    "similarityBoost": 0.75,
    "speed": 1.0
  },
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2",
    "language": "en-US",
    "smartFormat": true,
    "punctuate": true
  },
  "firstMessage": "Thank you for calling Bella Hair Studio! This is your virtual receptionist. How can I help you today?",
  "endCallMessage": "Thanks for calling Bella Hair Studio. We look forward to seeing you soon. Have a wonderful day!",
  "endCallPhrases": ["goodbye", "bye", "that's all", "thank you bye", "talk to you later"],
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "check_availability",
        "description": "Check available appointment slots for a given date and optionally a specific stylist",
        "parameters": {
          "type": "object",
          "properties": {
            "date": {
              "type": "string",
              "description": "Date in YYYY-MM-DD format"
            },
            "stylist": {
              "type": "string",
              "description": "Stylist name, or 'any' if caller has no preference"
            },
            "service_type": {
              "type": "string",
              "description": "Type of service requested (e.g. balayage, haircut, color)"
            }
          },
          "required": ["date", "service_type"]
        }
      },
      "server": {
        "url": "https://yourdomain.com/webhook/vapi",
        "timeoutSeconds": 10
      }
    },
    {
      "type": "function",
      "function": {
        "name": "create_booking",
        "description": "Create a confirmed appointment booking for a client after verifying availability",
        "parameters": {
          "type": "object",
          "properties": {
            "client_name": {
              "type": "string",
              "description": "Full name of the client"
            },
            "client_phone": {
              "type": "string",
              "description": "Client phone number including area code"
            },
            "date": {
              "type": "string",
              "description": "Appointment date in YYYY-MM-DD format"
            },
            "time_slot": {
              "type": "string",
              "description": "Appointment time, e.g. 10:00 AM"
            },
            "stylist": {
              "type": "string",
              "description": "Stylist name for the appointment"
            },
            "service_type": {
              "type": "string",
              "description": "Service being booked"
            }
          },
          "required": ["client_name", "client_phone", "date", "time_slot", "service_type"]
        }
      },
      "server": {
        "url": "https://yourdomain.com/webhook/vapi",
        "timeoutSeconds": 10
      }
    }
  ],
  "maxDurationSeconds": 600,
  "backgroundSound": "off",
  "backchannelingEnabled": true,
  "backgroundDenoisingEnabled": true,
  "silenceTimeoutSeconds": 30
}

Step 3: Write Your System Prompt Carefully

1. Business identity: Name, address, exact hours — including every exception (closed Tuesdays, different Sunday hours, holiday closures)
2. Complete service menu with prices: The AI should never say "I am not sure about pricing" — callers ask this on almost every call
3. Staff names and specialties: Callers often ask for specific people by name or by specialty
4. Booking rules: Deposit requirements, cancellation policy, how far in advance you book
5. Fallback instruction: What the AI should say when a question falls outside its knowledge
6. Tone guidance: "Warm and casual like a friendly coworker" vs. "Professional but approachable" produces noticeably different output

• Vague instructions like "be helpful" (too generic to guide behavior)
• Leaving out your cancellation policy (callers will ask about it)
• Forgetting to tell the AI it must use the check_availability tool rather than guessing

Step 4: Get a Phone Number and Set Up Call Routing

# AT&T — forward on no answer
*61*+1XXXXXXXXXX#

# Verizon — forward on no answer
*71XXXXXXXXXX

# T-Mobile — forward on no answer
**61*+1XXXXXXXXXX**

# Replace XXXXXXXXXX with your 10-digit Vapi number
# Dial from your business phone to activate; dial the deactivation code to disable

Step 5: Build the Webhook Server

from fastapi import FastAPI, Request
from fastapi.responses import JSONResponse
import sqlite3
from datetime import datetime
import httpx
import os

app = FastAPI()

DB_PATH = "/var/data/salon_bookings.db"

def get_db():
    conn = sqlite3.connect(DB_PATH)
    conn.row_factory = sqlite3.Row
    return conn

@app.post("/webhook/vapi")
async def vapi_webhook(request: Request):
    body = await request.json()
    message = body.get("message", {})
    msg_type = message.get("type")

    if msg_type == "function-call":
        function_name = message["functionCall"]["name"]
        params = message["functionCall"]["parameters"]

        if function_name == "check_availability":
            return await handle_check_availability(params)
        elif function_name == "create_booking":
            return await handle_create_booking(params)

    # Handle other Vapi message types (call started, ended, transcript, etc.)
    if msg_type == "end-of-call-report":
        await log_call_summary(message)

    return JSONResponse({"status": "ok"})

async def handle_check_availability(params: dict):
    date_str = params.get("date")
    stylist = params.get("stylist", "any")
    service = params.get("service_type", "")

    db = get_db()
    try:
        if stylist and stylist.lower() != "any":
            slots = db.execute("""
                SELECT time_slot, stylist_name
                FROM available_slots
                WHERE date = ? AND stylist_name LIKE ? AND is_booked = 0
                ORDER BY time_slot
                LIMIT 5
            """, [date_str, f"%{stylist}%"]).fetchall()
        else:
            slots = db.execute("""
                SELECT time_slot, stylist_name
                FROM available_slots
                WHERE date = ? AND is_booked = 0
                ORDER BY time_slot
                LIMIT 5
            """, [date_str]).fetchall()

        if not slots:
            return JSONResponse({
                "result": f"I don't see any openings on {date_str} for that service. Would you like me to check a different date, or I can put you on our cancellation list?"
            })

        slot_list = ", ".join([
            f"{row['time_slot']} with {row['stylist_name']}"
            for row in slots
        ])
        return JSONResponse({
            "result": f"Great news — on {date_str} we have: {slot_list}. Which of those works best for you?"
        })
    finally:
        db.close()

async def handle_create_booking(params: dict):
    db = get_db()
    try:
        db.execute("""
            INSERT INTO bookings
              (client_name, client_phone, date, time_slot, stylist, service, created_at, source)
            VALUES (?, ?, ?, ?, ?, ?, ?, 'vapi_phone')
        """, [
            params["client_name"],
            params["client_phone"],
            params["date"],
            params["time_slot"],
            params.get("stylist", "TBD"),
            params["service_type"],
            datetime.utcnow().isoformat()
        ])

        # Mark the slot as booked so it cannot be double-booked
        db.execute("""
            UPDATE available_slots
            SET is_booked = 1
            WHERE date = ? AND time_slot = ? AND stylist_name = ?
        """, [params["date"], params["time_slot"], params.get("stylist", "TBD")])

        db.commit()

        # Send SMS confirmation via Twilio
        await send_sms_confirmation(params)

        return JSONResponse({
            "result": (
                f"You're all set, {params['client_name']}! "
                f"I've booked your {params['service_type']} appointment on {params['date']} "
                f"at {params['time_slot']}. "
                f"You'll receive a confirmation text at {params['client_phone']} shortly."
            )
        })
    except Exception as e:
        db.rollback()
        return JSONResponse({
            "result": (
                "I ran into a small issue completing that booking on my end. "
                "Let me have one of our team members call you back within the hour to confirm. "
                "Can I verify I have the right number for you?"
            )
        })
    finally:
        db.close()

async def send_sms_confirmation(booking: dict):
    """Send an SMS booking confirmation via Twilio."""
    async with httpx.AsyncClient() as client:
        response = await client.post(
            f"https://api.twilio.com/2010-04-01/Accounts/YOUR_TWILIO_ACCOUNT_SID/Messages.json",
            auth=("YOUR_TWILIO_ACCOUNT_SID", "YOUR_TWILIO_AUTH_TOKEN"),
            data={
                "From": "+1YOUR_TWILIO_PHONE_NUMBER",
                "To": booking["client_phone"],
                "Body": (
                    f"Hi {booking['client_name']}! Your {booking['service_type']} appointment "
                    f"at Bella Hair Studio is confirmed for {booking['date']} at {booking['time_slot']}. "
                    f"Reply CANCEL to cancel (24hr notice required). See you soon!"
                )
            }
        )
    return response

async def log_call_summary(message: dict):
    """Log call metadata for analytics."""
    db = get_db()
    try:
        db.execute("""
            INSERT INTO call_logs (call_id, duration_seconds, ended_reason, transcript_summary, created_at)
            VALUES (?, ?, ?, ?, ?)
        """, [
            message.get("call", {}).get("id"),
            message.get("durationSeconds", 0),
            message.get("endedReason"),
            message.get("summary", ""),
            datetime.utcnow().isoformat()
        ])
        db.commit()
    finally:
        db.close()

A Real Conversation: What Callers Actually Hear

Vapi: Thank you for calling Bella Hair Studio! This is your virtual receptionist. How can I help you today?

Caller: Hi yeah, I wanted to book — actually, do you guys do balayage?

Vapi: We absolutely do! Our color specialist Maria does beautiful balayage work. Prices start at $180 depending on your hair length and thickness. Were you thinking of booking an appointment?

Caller: Yeah, how long does it take? I have a thing in the afternoon.

Vapi: A balayage service typically takes 2.5 to 3 hours, so you would want to plan for a morning slot. What date were you thinking?

Caller: Maybe this Saturday?

Vapi: Let me check Saturday for you. (1.8-second pause while calling check_availability) We have Maria available at 10 AM and at 11 AM this Saturday. Do either of those work?

Caller: 10 AM works. My name is Jessica — wait, actually can I do 11? I forgot I have something at 10.

Vapi: Of course, 11 AM is perfect. Can I get your last name and the best phone number to reach you?

Caller: Jessica Torres. Number is 512-555-0847.

Vapi: Got it. (2-second pause while creating the booking) You are all set, Jessica! Your balayage appointment with Maria is confirmed for this Saturday at 11 AM. You will receive a confirmation text at 512-555-0847 shortly. Anything else I can help with?

Caller: No, that is great. Thank you!

Vapi: Wonderful — we look forward to seeing you Saturday. Have a great day!

Cost Reality Check

Vapi vs. The Competition

Common Mistakes to Avoid

FAQ

Never Miss a Call Again

Related Articles

• Bland AI for Local Business: Automated Phone Answering That Sounds Human
• Retell AI vs Vapi for Local Business: Which Phone AI Wins in 2026?
• ElevenLabs Voice AI for Local Business Customer Calls
• AI Receptionist for Small Business: Complete Setup Guide 2026
• Claude MCP + Google Calendar: Booking Automation for Local Business
• AI Phone Answering Agent for Local Business