AI voice agents that sound like a person.

A conversation, not a phone menu.

A voice agent is software that holds a spoken conversation. The caller talks normally; the agent works out what they want, answers out loud and can act on it: book the appointment, look up the order, update the CRM.

Ours runs on speech models we build. Psst cleans the audio, speech recognition turns it into text, Turno decides when the caller has finished or is interrupting, a language model writes the reply and Versa speaks it, streaming as it is written.

• InterruptionsCallers can talk over the agent. Turno tells a real interruption from a “mhm”.
• ToolsYour own functions, calendar booking, CRM lookups and web search, during the call.
• KnowledgeDocuments and links the agent answers from, and memory of earlier calls with the same caller.
• Phone and webLocal numbers in the US, Spain, the UK, Germany, France, Italy and Portugal, inbound and outbound, or our SDKs in your website and app. Phone agents
• LanguagesAgents listen and answer in Spanish, English, Portuguese, French, Italian, German. Catalan, Galician, Basque: the voice speaks them, agents can’t listen in them yet.
• Price4.2¢ a minute for the platform, billed per second, plus the language model per token: about 5.2¢ a minute at a typical prompt. $5 of credit to start.

A voice agent in ten lines.

import { VoiceAgentClient } from '@lokutor/sdk';

const agent = new VoiceAgentClient({
  apiKey: 'YOUR_API_KEY',
  prompt: 'You are the receptionist for a dental clinic.',
  voice: 'F2',
  language: 'es',
  serverUrl: 'wss://api.lokutor.com/ws/agent',
});

await agent.startManaged();

Python SDK, REST and WebSocket APIs, a Pipecat service and an MCP server are in the documentation.