Skip to main content

​
What is Retone AI?

Retone AI is a voice AI agent platform that enables businesses to build, deploy, and manage sophisticated AI-powered phone agents. These agents can handle inbound and outbound calls, engage in natural conversations, collect information, book appointments, and seamlessly transfer to human agents when needed.

​
Core Concepts

​
Flow Agents

Flow Agents are AI-powered entities that handle your phone conversations using a visual, node-based design. Build sophisticated conversation flows with:
  • Multi-step processes - Guide callers through complex workflows
  • Data collection - Gather and validate customer information
  • Conditional routing - Branch conversations based on user input
  • Appointment booking - Schedule meetings directly in calls
  • Human transfer - Seamlessly escalate to live agents

​
The AI Stack

Retone AI integrates best-in-class providers for each component of the voice AI pipeline:
ComponentProvidersPurpose
Speech-to-Text (STT)Deepgram Flux, AssemblyAIConvert caller speech to text in real-time
Language Model (LLM)Groq, OpenAI, AnthropicGenerate intelligent, contextual responses
Text-to-Speech (TTS)Cartesia Sonic 3, ElevenLabsConvert agent responses to natural speech
TelephonyTwilio, LiveKit SIPHandle phone connections and call routing

​
Flow Builder

The Flow Builder is a visual canvas for designing conversation flows. Instead of writing one large prompt, you create a graph of nodes connected by transitions. 
[Start] → [Greeting] → [Collect Info] → [If/Else] → [Book] or [Transfer]
                                            ↓
                                        [End Call]
Node Types:
  • Start - Entry point for conversations
  • Conversation - AI-powered dialogue stages
  • Data Extraction - Collect and validate information
  • If/Else - Branch based on conditions
  • Calendar Booking - Schedule appointments
  • Human Transfer - Connect to live agents
  • Agent Transfer - Hand off to another AI agent
  • Function - Execute custom backend logic
  • End Call - Terminate the conversation

​
Knowledge Base

Power your agents with your own data using RAG (Retrieval-Augmented Generation):
  • Upload Documents - PDF, TXT, DOCX files
  • Scrape Websites - Import content from web pages
  • Organize in Folders - Group related content
  • Link to Agents - Attach specific knowledge to agents or flow nodes
When a caller asks a question, the agent searches your knowledge base and incorporates relevant information into its response.

​
Architecture Overview

                    ┌─────────────────────────────────────────┐
                    │              Phone Call                  │
                    └─────────────────┬───────────────────────┘
                                      │
                    ┌─────────────────▼───────────────────────┐
                    │         Telephony (SIP/Twilio)          │
                    └─────────────────┬───────────────────────┘
                                      │
          ┌───────────────────────────┼───────────────────────────┐
          │                           │                           │
          ▼                           ▼                           ▼
    ┌───────────┐             ┌───────────────┐            ┌───────────┐
    │    STT    │             │  Voice Agent  │            │    TTS    │
    │ (Deepgram)│────────────▶│   (LLM +      │───────────▶│(Cartesia) │
    └───────────┘             │  Flow Logic)  │            └───────────┘
                              └───────┬───────┘
                                      │
                    ┌─────────────────┼─────────────────┐
                    │                 │                 │
                    ▼                 ▼                 ▼
             ┌───────────┐    ┌───────────┐     ┌───────────┐
             │ Knowledge │    │  Calendar │     │  Actions  │
             │   Base    │    │  System   │     │ (Transfer)│
             └───────────┘    └───────────┘     └───────────┘

​
Key Features

​
Telephony

  • Provision phone numbers directly in the platform
  • Import existing Twilio or SIP trunk numbers
  • Handle inbound and outbound calls
  • Call recording with configurable retention

​
Call Transfers

  • Warm Transfer - Brief the human agent before connecting
  • Time Restrictions - Route based on business hours
  • Fallback Handling - Graceful handling when transfers unavailable

​
Data Extraction

  • Collect structured information during calls
  • Validate with regex patterns (email, phone, custom formats)
  • Support for multiple data types (text, number, date, boolean, select)

​
Analytics

  • Call volume and duration metrics
  • Success rate tracking
  • Sentiment analysis
  • Peak hours identification
  • Export capabilities

​
Security

  • Encrypted SIP credentials
  • Organization-level isolation
  • Role-based access control
  • Call recording controls

​
Getting Started

Ready to build your first voice agent?

Quick Start

Create an agent in 5 minutes

Flow Builder Guide

Learn visual conversation design