4. Create Voice Agent
5 min
This section walks you through creating and configuring a Voice Agent end-to-end.
What you’ll finish with
- A Voice Agent with a clear name and description
- Voice configuration completed (STT, LLM, and TTS providers)
- Workflows connected (only if your agent needs to take actions/tools)
- Tasks written (your main prompts/instructions)
- Memory configured (optional)
- Final review completed and the agent created
Before you start (recommended)
- Define the goal (1 sentence): clearly define what the agent should do
- Example: “Help users book hospital appointments through a voice conversation.”
- Collect reference info: product details, policies, FAQs, tone guidelines
- Voice provider readiness: ensure required providers are configured:
- Speech-to-Text (STT)
- LLM (AI model)
- Text-to-Speech (TTS)
- Workflow readiness: confirm required workflows/paths are already created and tested
- Attribute clarity: keep attribute names consistent (same spelling, meaning, and format)
- Keep it minimal: remove unused configurations to avoid confusion
Quick overview (follow in order)
Follow the steps below to configure and create a Voice Agent:
- Step 1 — Navigate to Agents
- Step 2 — Start Agent Creation
- Step 3 — Define Agent (Voice Configuration)
- Step 4 — Workflow Setup
- Step 5 — Add Tasks (Prompt Writing)
- Step 6 — Set Up Memory (Optional)
- Step 7 — Evaluate / Disposition Setup & Review
Video: Creating a Voice Agent
Tips for better results
- Keep the agent goal clear and focused
- Ensure responses are short and natural for voice conversations
- In Tasks, define:
- Tone (friendly, professional, calm)
- Speaking style (simple, conversational)
- Do/Don’t rules
- When to use workflows
- Avoid long or complex responses (voice should be easy to understand)
- Add 1–2 examples (optional) to improve consistency
- If responses are incorrect, update the Task prompt first before adding more workflows