How to build a voice agent

Definition

Building a voice agent used to mean wiring together a speech-to-text service, a language model, a text-to-speech service, and a telephony provider, then keeping all four in sync. With an all-in Voice Agent API, it is three steps: get a key, create the agent, connect it to a surface.

Step 1: get a key

Sign up and generate an API key from the dashboard. Set it as SPEECHIFY_API_KEY so the SDKs pick it up. 60 minutes are free every month with commercial use, so you can test with real calls before paying anything.

Step 2: create the agent

Create an agent with one POST to /v1/agents, setting a prompt, a voice, a first message, and a language. The Python (speechify-api) and TypeScript (@speechify/api) SDKs wrap the same call. That single request gives you a working agent.

Step 3: connect a surface

Point the agent at a surface. Provision a phone number or connect your own SIP trunk for calls, or run the agent over WebRTC in a browser. The same agent configuration drives phone and web, so you build once and deploy to both.

Next steps

From a basic agent, add tools so it can act on your systems, a knowledge base so it answers from your content, and testing so regressions surface before a real caller hits them. Each is incremental on top of the three steps.