Skip to main content
Lyzr Studio’s Voice Agents allow you to build sophisticated, conversational AI agents capable of handling real-time voice interactions. These agents can be deployed via phone numbers (Twilio) and are equipped with RAG capabilities, tool calling via sub-agents, and comprehensive monitoring.

​
accessing Voice Agents

To get started, navigate to the Voice section in the main sidebar. Upon entering the Voice module, you will see the dashboard. Here you can manage your existing agents, configure phone numbers, and monitor active calls. To build a new agent, click the + Create Agent or + Create new button.

​
Configuring Your Agent

The Create Agent screen allows you to define the core identity and behavior of your voice bot.

​
Basic Settings

  • Name & Description: meaningful identifiers for your agent.
  • LLM Provider & Model: Select the reasoning model (e.g., OpenAI GPT-4o) that powers the agent’s intelligence.
  • Who Speaks First: Choose whether the Human Speaks First (listening mode) or the Agent Speaks First (greeting mode) when a call connects.

​
Language Support

Lyzr Voice Agents support multiple languages. Select your target language from the dropdown menu to ensure the agent speaks and understands the correct dialect.

​
Voice Personality

You can select a specific voice profile (e.g., Sage, Alloy, Ash) to match the tone of your brand or use case.

​
Defining Persona & Instructions

The core logic of your agent is defined by its Role, Goal, and Instructions. You can write these manually or use the Generate with AI feature to create optimized system prompts.
  • Agent Role: Defines the persona (e.g., “Compassionate front-desk coordinator”).
  • Agent Goal: The primary objective of the call (e.g., “Screen incoming calls for dental emergencies”).
  • Agent Instructions: detailed, step-by-step behavioral guidelines. This is where you define how the agent handles greetings, triage logic, and fallbacks.

​
Knowledge Base (RAG)

To make your Voice Agent context-aware using your own data, toggle the Knowledge Base switch. Configure the Retrieval-Augmented Generation (RAG) settings:
  1. Select Knowledge Base: Choose a pre-uploaded corpus of documents.
  2. Number of Chunks: Define how many text segments the AI should retrieve per query.
  3. Retrieval Type: Choose the search strategy (e.g., Basic).
  4. Score Threshold: Set the minimum relevance score for retrieved data to be considered valid.

​
Manager Agent & Tool Calling

To enable your Voice Agent to perform actions (like booking appointments, sending emails, or updating CRMs), you must enable the Manager Agent toggle. How it works:
  • Voice Agents typically do not call tools directly.
  • Instead, you assign a Sub-Agent (created in the Agents module) specifically designed to handle tool execution.
  • The Manager Agent (the Voice Agent) detects when a tool is needed and routes that specific part of the request to the Sub-Agent to execute the task.