v1.2.0 is out MIT License

Build voice agents
that actually listen.

The open-source orchestration engine for production-grade AI voice agents. Low latency, provider agnostic, and built for scale.

npm install sharyx-os Star on GitHub 
import { createAgent, OpenAILLM, DeepgramSTT, ElevenLabsTTS } from 'sharyx-os';

const agent = createAgent({
  stt: new DeepgramSTT({ apiKey: process.env.DEEPGRAM_API_KEY }),
  llm: new OpenAILLM({ apiKey: process.env.OPENAI_API_KEY }),
  tts: new ElevenLabsTTS({ apiKey: process.env.EL_API_KEY }),
});

agent.start({ port: 3000 });
Core Capabilities

Everything you need for voice AI

⚡ Low Latency

Optimized streaming architecture for near-instant responses. Sub-500ms conversational loops.

🔄 Tool Calling

Enable agents to search, book, and act in real-time with recursive tool calling support.

🔌 Provider Agnostic

Swap between OpenAI, Gemini, Deepgram, and ElevenLabs with one line of code.

📞 Multi-Channel

Native support for Twilio, Plivo, and WebRTC. Deploy your agent anywhere.

🧠 Persistent Memory

Redis-backed session management for stateful, long-running conversations.

🛠️ CRM Integrations

Built-in support for HubSpot, Google Calendar, and WhatsApp Cloud API.

How it works

The Orchestration Engine

User
Transport
STT (Deepgram)
Sharyx OS
LLM (GPT-4)
TTS (ElevenLabs)
Easy Setup

Implementation Procedures

Simple steps to get your voice agent up and running.

📞

Telephony Procedure

1

Install with npm install sharyx-os

2

Add Provider keys to your .env

3

Launch the server via npm run start

4

Connect your Twilio/Plivo webhook

🌐

Webcall Procedure

1

Install with npm install sharyx-os

2

Add your API keys to the .env file

3

Run the environment with npm run start

4

Open your browser to start calling

Ready to start building?

Get started with our lightweight SDK today.

npm install sharyx-os