Voice agents
A voice agent is a NappAI flow that holds spoken conversations. The person speaks, the agent listens, understands what they are asking for, uses the flow’s tools if it needs them and answers with a voice. It can answer calls from your website or from a phone, and it can also make the call itself.
The pieces
Every voice agent is a flow with a Voice Agent component and, connected to it:
| Connection | What it adds |
|---|---|
| Model | The AI model that leads the conversation and decides when to use a tool. |
| STT | The ears: turns what the person says into text. |
| TTS | The voice: turns the agent’s answer into audio. |
| Tools | What the agent can do during the call: look up data, book an appointment, transfer the call, hang up… |
Test and publish
As with any flow, the playground tests the draft: it opens a test call with what is on the canvas. Real calls use the published version.
The full guide
The Voice Agents section explains each step with screenshots:
- How a voice agent works
- Create a voice agent
- Test in the playground
- The Voice Agent menu and publishing
- The avatar
- The agent on your website
- Phone: inbound calls
- Outbound calls
- Transfers and hanging up
- History, transcripts and recordings
- Quality, noise and latency
- Troubleshooting
The settings of each field are on the Voice Agent component page.