Browser Agent SDK is live: four packages, one install

[naomi-deepgram]

We just shipped the Browser Agent SDK on npm:

• @deepgram/agents-widget: drop-in widget, six layouts, no framework required
• @deepgram/ui: pre-built React components, CSS-variable themed
• @deepgram/react: provider + hooks for state, audio, and function calling
• @deepgram/agents: framework-agnostic core. WebSocket client, microphone capture, player

Each layer builds on the one below it. Install the highest layer you need and everything beneath comes with it. All layers share reconnection (exponential backoff + jitter), playback-aware mode tracking, audio buffering, optional Silero VAD, KeepAlive pings, and a typed event emitter for server messages.

Auth uses a tokenFactory pattern: your server mints short-lived tokens via the Deepgram auth grant endpoint, the SDK calls it on every connect and reconnect, and your API key never touches the browser. The token rides Sec-WebSocket-Protocol because that is the only custom header browsers permit on WebSocket handshakes.

Quickstart

npm install @deepgram/agents-widget

import { init } from "@deepgram/agents-widget";

init({
  tokenFactory: () => fetch("/api/deepgram-token").then((r) => r.text()),
  agent: "YOUR_AGENT_ID",
  layout: "floating",
});

• Full overview: developers.deepgram.com/docs/browser-agent-overview
• Blog: deepgram.com/learn/browser-agent-sdk

Feedback and questions welcome.