Vapi

Voice AI Platform for Building Conversational Agents

What is Vapi?

Vapi is a developer-first platform for building, testing, and deploying AI-powered voice assistants. It provides a complete voice AI stack—handling telephony, speech-to-text, LLM orchestration, and text-to-speech—so developers can focus on building conversational experiences without managing complex audio pipelines.

"Vapi makes it easy to build voice AI agents that can make and receive phone calls, with sub-500ms latency and seamless integration with popular LLMs like GPT-4, Claude, and open-source models."

— Vapi.ai

Telephony

Inbound & outbound

Low Latency

<500ms response

Multi-LLM

GPT-4, Claude, etc.

Function Calling

Custom tools

How Vapi Works

Phone Call STT LLM TTS Audio Out

Vapi abstracts away the complexity of real-time voice AI, handling audio streaming, turn-taking, interruption handling, and latency optimization automatically.

Core Features

Assistant Configuration

Define your voice assistant with a system prompt, voice selection, LLM provider, and conversation settings. Vapi handles the orchestration.

System Prompt Voice Selection LLM Provider First Message

Telephony Integration

Built-in phone number provisioning and SIP trunking. Make outbound calls or receive inbound calls with automatic call handling.

Inbound Calls Outbound Calls SIP Trunking WebRTC

Function Calling / Tools

Connect your assistant to external APIs and databases. Define custom functions that the LLM can invoke during conversation—booking appointments, querying data, etc.

Custom Functions Webhooks API Integration

Interruption Handling

Natural barge-in support. When users interrupt, Vapi stops speaking and processes the new input—creating natural, human-like conversations.

Barge-in VAD Turn-taking

Supported Providers

LLM Providers

• OpenAI (GPT-4, GPT-4o)
• Anthropic (Claude 3)
• Groq (Llama, Mixtral)
• Together AI
• Custom (OpenAI-compatible)

STT Providers

• Deepgram
• AssemblyAI
• Gladia
• Whisper (OpenAI)

TTS Providers

• ElevenLabs
• OpenAI TTS
• PlayHT
• Deepgram
• Rime

Quick Start

# Create an outbound call with Vapi
import requests

response = requests.post(
    "https://api.vapi.ai/call/phone",
    headers={
        "Authorization": "Bearer YOUR_API_KEY",
        "Content-Type": "application/json"
    },
    json={
        "assistantId": "your-assistant-id",
        "phoneNumberId": "your-phone-id",
        "customer": {
            "number": "+15551234567"
        }
    }
)

print(response.json())

// Assistant configuration example
{
    "name": "Customer Support Agent",
    "model": {
        "provider": "openai",
        "model": "gpt-4o",
        "systemPrompt": "You are a helpful customer support agent..."
    },
    "voice": {
        "provider": "elevenlabs",
        "voiceId": "rachel"
    },
    "firstMessage": "Hello! How can I help you today?"
}

Tip: Use the Vapi Dashboard to create and test assistants without code, then integrate via API for production.

Use Cases

Customer Support

24/7 AI phone support

Appointment Booking

Automated scheduling

Sales Outreach

Lead qualification calls

Surveys

Automated phone surveys

Reminders

Appointment confirmations

Healthcare

Patient intake & triage

Resources & References

Documentation

docs.vapi.ai

Dashboard

Create & test assistants

GitHub

SDKs & examples

Discord Community

Get help & share