livekit/livekit
End-to-end realtime stack for connecting humans and AI
What it solves
LiveKit provides the underlying infrastructure needed to build real-time audio and video applications. It solves the complexity of moving high-bandwidth media streams between people, devices, and AI agents with low latency and high scalability.
How it works
At its core, this repository is a scalable, distributed WebRTC Selective Forwarding Unit (SFU) written in Go. It acts as a central hub that routes real-time audio, video, and data tracks between participants in a virtual room. AI agents are treated as first-class participants, joining rooms just like human users, and can be automatically routed into sessions via an agent dispatch system.
Who it’s for
It is designed for developers building voice AI assistants, video conferencing tools, real-time teleoperation systems, or any application requiring synchronized, low-latency media streaming across multiple platforms.
Highlights
- AI Agent Integration: Specifically built to support AI agents as room participants with dedicated dispatching logic.
- Scalable Infrastructure: Distributed SFU architecture that supports multi-region deployments and high-volume traffic.
- Broad Compatibility: Provides SDKs for web, mobile, desktop, embedded devices, and servers.
- Advanced Media Features: Supports simulcast, selective subscription, end-to-end encryption, and SVC codecs (VP9, AV1).
- Versatile Connectivity: Handles UDP, TCP, TURN, and telephony over SIP.
Related
- Project
- Project
- Project
- Project
- Project