AI Connect OSAI Connect OS Documentation
DocumentationAPI Reference
Voice Agents

Voice Agents

Build AI agents that make and receive phone calls with natural conversation.

Voice docs are being rewritten. The dashboard's Agent Builder is the fastest way to get an agent live today.

Concepts

  • Agent - a configured voice persona with a system prompt, voice, tools, knowledge bases, and call-handling settings.
  • Phone number - a Twilio, Telnyx, or Vonage number bound to one or more agents for inbound and outbound calls.
  • Call - a single connection between a phone number and a caller, with transcript, recording, summary, and structured outputs.
  • Tool - a function your agent can call mid-conversation (book an appointment, look up an order, transfer the call).
  • Knowledge base - a collection of uploaded documents the agent retrieves from when answering questions.

Provider stack

Voice agents are built on a pluggable provider model so you can pick the best engine for each layer of the call:

  • Speech-to-text (STT) - Deepgram (default), Deepgram Flux (beta), AssemblyAI, OpenAI Whisper
  • Text-to-speech (TTS) - ElevenLabs (default), Cartesia, OpenAI, Azure
  • LLM - Anthropic Claude, OpenAI, Groq, Cerebras
  • Telephony - Twilio, Telnyx, Vonage (per-number credentials)

Next steps

  • Open the dashboard's Voice → Agents section to create an agent.
  • Buy or port a phone number under Voice → Phone Numbers.
  • See the Authentication page for API-driven agent management.

On this page