Setup a simple Voice Agent
A complete, end-to-end guide to designing, configuring, and operating intelligent phone-based AI agents.
1. Introduction
Angoor AI enables organizations to build, deploy, and manage intelligent voice agents capable of conducting real-time conversations over phone calls. These voice agents understand spoken language, process user intent, access business information, make decisions, and respond naturally using AI-generated speech.
This document is the complete reference for building such an agent. It is written so that a business stakeholder can understand what each part does and why it matters, while an engineer or solution architect can configure the platform directly from these instructions.
1.1 What You Will Build
By the end of this guide, you will have a production-ready voice agent that can answer a phone call, greet the caller, hold a natural conversation, look up and act on business information, and gracefully end the call when the conversation is genuinely complete, all without a human operator.
1.2 The Four Building Blocks
A complete voice agent within Angoor AI is assembled from four primary components. Each has one clear job, and together they form the full conversational workflow.
Plivo Integration — Handles telephony, speech recognition (STT), and voice generation (TTS).
Agent Configuration — Defines the intelligence, personality, and behaviour of the AI.
Tools — Provide specialized capabilities such as classification, routing, extraction, and orchestration.
Flows — Control the sequence of actions and decisions performed during a conversation.
Together, these components handle customer support, lead qualification, appointment scheduling, information retrieval, and many other business use cases.
2. How Angoor Voice Agents Work (STT-TTS Integration)
Every voice conversation follows the same real-time pipeline. When a caller speaks, their audio is converted into text by a speech-to-text (STT) provider. The Agent processes that text using its configured prompt, tools, variables, and business rules. The generated response is converted back into speech by a text-to-speech (TTS) provider and delivered to the caller — all within a fraction of a second, and repeated on every turn until the call ends.
Read the diagram left to right: the caller enters through the Plivo telephony layer -> audio is transcribed by STT -> the Agent reasons over the text and may call Tools -> the reply is synthesized by TTS and played back to the caller.
💡 Why this matters for the business Each component can be swapped or tuned independently. You can change voice providers, upgrade the model, or adjust call-ending behavior without rebuilding the agent — which keeps long-term maintenance cost low.
3. Prerequisites
Before configuring a voice agent, ensure the following are available:
An active Angoor AI workspace.
A configured Plivo account (for telephony).
Access to STT and TTS providers.
Agent configuration permissions.
Flow builder access.
The required business prompts and workflows for your use case.
Navigation note:
Throughout this guide, configuration paths are shown as Configuration → Section. The left-hand navigation in Angoor AI contains: Agents, Flows, Channel Management, Forms, Entities, Tools, Integrations, and (under Deployments) Widget.
Last updated