Get started / Getting started

Build your first voice agent

From zero to a live agent answering calls in four steps — set your keys, pick your models, write the agent, and dial in.

Quickstart

Four steps to your first call.

1. 1

   Set your API keys

   Copy .env.example to .env and fill in the required keys — Clerk, Postgres, LiveKit, and your model providers.

2. 2

   Pick your models

   Choose the STT, LLM, and TTS. Defaults work out of the box; override per-project or per-agent.

3. 3

   Write the agent

   One import, one file. The SDK handles voice, telephony, and routing.

4. 4

   Make a call

   Attach a number to your agent and dial in — it answers and runs the process.

Write the agent SDK

Install @voicelayer/sdk — then this is a complete agent:

import { Agent } from '@voicelayer/sdk';

export default await new Agent({
  name: 'fnol-simple',
  prompt: 'You are a calm First Notice of Loss intake agent.',
  capabilities: ['fnol', 'inbound-sip'],
  tools: {
    transferToHuman: {
      description: 'Transfer the caller to a human claims agent.',
      run: async (_, ctx) => ctx.handoff('+1 555 012 3456'),
    },
  },
}).start(import.meta.url);

Your default models

Every call runs a STT → LLM → TTS pipeline. These are the defaults — override per-project (BYOK) or per-agent.