Overview
Flow Builder supports voice deployment through Telnyx. Connect a phone number to your flow so inbound callers interact with your AI agent over a live phone call with real-time speech processing.How It Works
- Telnyx receives the inbound call and forwards it to your flow’s voice endpoint
- Flow Builder processes each turn: speech-to-text → node execution → LLM call → response
- Responses stream back via SSE for low-latency text-to-speech playback
- Conversation continues until an End node is reached or the caller hangs up
Setup
Create a Telnyx Account
Sign up at telnyx.com and purchase a phone number.
Configure the Webhook
In Telnyx, set the phone number’s webhook URL to your flow’s voice endpoint (found in Flow Builder’s deploy settings). Set the method to POST.
Voice-Specific Best Practices
| Consideration | Recommendation |
|---|---|
| Model selection | Use fast models (gpt-4o-mini, claude-haiku) for low latency |
| Prompt length | Keep prompts concise to reduce processing time |
| Static messages | Use static mode for greetings/disclaimers to skip LLM |
| Tool calls | Enable speak-during-execution filler to avoid dead air |
Troubleshooting
| Issue | Solution |
|---|---|
| Calls connect but no audio | Verify webhook URL matches your flow’s voice endpoint |
| Long pauses between turns | Use a faster LLM model and shorten prompts |
| Agent doesn’t respond | Confirm API key is configured |
| Call drops immediately | Check Telnyx webhook logs for connection errors |
Related Pages
Engine & Streaming
How streaming works for voice
Testing
Test before deploying
Embedding Widget
Deploy to chat instead
API Key Setup
Required for voice