

Two ways to build
Autopilot
Fonoster runs the agent for you. Write the prompt, pick the voice and the language model, and
give it tools. You don’t run a server.
Streams
You run the agent on your own WebSocket server. Write it with the Voice SDK for Node.js or
Python, or connect a Twilio or Pipecat bot you already have.
What every call gets
Both ways share the same phone layer, so you don’t build it yourself:- Numbers and SIP trunks from the carrier you choose
- Transfers to a person, cold or warm
- DTMF: read and send keypad tones
- Answering-machine detection, most useful on outbound calls
- Recording, with a consent prompt when you need one
- A signed record of every call, with its transcript and recording
- Webhooks when calls start and end
Start here
Quickstart
Create an Autopilot agent and call it.
How Fonoster works
Learn how workspaces, numbers and applications fit together.