Quickstart: in the browser

Talk to an agent from a web page in about five minutes.

You need an API key from API Keys in the dashboard. Keep it in your environment as OMNIA_API_KEY.

1. Pick a language and voice

   curl "https://api.omnia-voice.com/api/v1/languages" -H "X-API-Key: $OMNIA_API_KEY"
   curl "https://api.omnia-voice.com/api/v1/voices?languageCode=en" -H "X-API-Key: $OMNIA_API_KEY"
   

   Keep one languageId and one voiceId. Each voice has a previewUrl worth listening to.

2. Create an agent

   curl -X POST "https://api.omnia-voice.com/api/v1/agents" \
     -H "X-API-Key: $OMNIA_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{
       "name": "First agent",
       "status": "ACTIVE",
       "config": {
         "languageId": "LANGUAGE_ID",
         "voiceId": "VOICE_ID",
         "greeting": "Hi! What can I help you with?",
         "basePrompt": "You are a friendly assistant. You are brief and never guess — if you do not know something, say so.",
         "firstSpeaker": "agent",
         "temperature": 0
       }
     }'
   

   Save the returned id.

   You write configuration under config and read it back under standardConfig. The request and response are not symmetric.

3. Create the call — on your server

   // POST /api/voice/start
   const res = await fetch("https://api.omnia-voice.com/api/v1/calls/create", {
     method: "POST",
     headers: {
       "X-API-Key": process.env.OMNIA_API_KEY,   // stays server-side
       "Content-Type": "application/json",
     },
     body: JSON.stringify({ agentId: AGENT_ID, connectionType: "webrtc" }),
   });
   
   const { websocketUrl } = await res.json();
   return Response.json({ websocketUrl });        // only this reaches the browser
   

   Do this server-side. An API key in browser JavaScript is readable by anyone who opens devtools, whatever bundler you use.

4. Connect from the browser

   npm install @omnia-voice/sdk
   

   import { OmniaSession } from "@omnia-voice/sdk";
   
   const session = new OmniaSession({ apiKey: PLACEHOLDER });
   
   session.addEventListener("transcripts", () => {
     console.log(session.transcripts.at(-1));
   });
   
   const { websocketUrl } = await (await fetch("/api/voice/start", { method: "POST" })).json();
   await session.joinCall({ websocketUrl });
   

   Start this from a real click — browsers deny microphone permission far more often on page load, and block audio playback until the user interacts.

Next

Give it tools

Let it act, not just talk.

Drive your UI

Client tools, with a worked example.