Voice Agents Overview

Deploy AI-powered phone agents that handle inbound calls, understand natural language, and take action.

What are Voice Agents?

Voice Agents are AI-powered phone assistants that can answer inbound calls, carry on natural conversations, and execute actions — all without human involvement. They combine:

• Speech recognition (transcriber) to convert caller audio to text
• A large language model (LLM) to understand intent and generate responses
• Text-to-speech (voice) to speak replies back to the caller
• Tools to take real-world actions (look up data, save contact attributes, send WhatsApp messages)
• Knowledge base for accurate, grounded answers to domain-specific questions

Use cases

Use case	Example
Inbound customer support	Answer product FAQs, troubleshoot issues, escalate complex cases
Lead qualification	Ask callers qualifying questions, save answers as contact attributes
Appointment booking	Collect booking details via voice and send a WhatsApp confirmation
Order status updates	Look up order status via HTTP tool and read it back to the caller
After-hours coverage	Handle calls 24/7 when human agents are offline

Call flow

Caller dials your Plivo phone number
         ↓
Swiftsell answers the call
         ↓
Voice agent starts (plays greeting or waits for caller, based on firstSpeaker setting)
         ↓
Caller speaks → transcriber converts speech to text
         ↓
LLM processes the text → generates a response or decides to call a tool
         ↓
If tool call: executes HTTP request / saves attribute / sends WhatsApp message
         ↓
Text-to-speech speaks the reply to the caller
         ↓
Conversation continues until caller hangs up or agent ends the call
         ↓
Call log created with duration, direction, and link to full transcript

Key concepts

Phone number — Each voice agent is linked to a Plivo phone number. When that number receives a call, the associated agent handles it. Configure phone numbers in Settings → Phone Number.

firstSpeaker — Determines who speaks first when a call connects. Set to agent to have the AI greet the caller immediately, or user to wait for the caller to speak first.

System prompt — The instructions that define the agent’s personality, scope, and behaviour. A well-written system prompt is the most important factor in agent quality.

Tools — Optional capabilities the agent can invoke mid-conversation (HTTP requests, attribute updates, WhatsApp messages). See Agent Tools.

Knowledge base — Attach a knowledge base to ground the agent’s answers in your own content. See Attach a Knowledge Base.

Next steps

Create a Voice Agent

Set up your first agent with a name, system prompt, and phone number.

Configure LLM & Voice

Choose your LLM provider, transcriber, and text-to-speech voice.

Add Tools

Give your agent the ability to call APIs, save attributes, and send messages.

View Call Logs

Review call history and access full conversation transcripts.