LiveKit

LiveKit is a WebRTC media server and surrounding platform for adding multi-user real-time video, audio, and data to applications. The server is a Go implementation of a selective forwarding unit (SFU) built on Pion WebRTC; it ships as a single binary, Docker image, or Helm chart, and can run as a distributed, multi-region deployment or on the hosted LiveKit Cloud.

Participants join rooms using JWT access tokens that encode identity and room permissions. Server SDKs generate those tokens, call the server APIs, and receive webhooks, while client SDKs cover web, mobile, and native runtimes. The companion Agents framework lets you attach programmable backend participants, which is how it serves as the transport layer for voice and multimodal AI agents, with automatic turn detection and interruption handling on the agent side.

Features

• Distributed SFU: scalable selective forwarding with UDP, TCP, and TURN connectivity
• Client SDKs: JavaScript/TypeScript (with React), Swift, Kotlin, Flutter, Unity WebGL, React Native, and Rust
• Server SDKs: Go, JavaScript/TypeScript, Ruby, Java/Kotlin, plus community Python and PHP
• JWT authentication: tokens carry participant identity and granted room permissions
• Media features: speaker detection, simulcast, selective subscription, SVC codecs (VP9, AV1), and end-to-end encryption
• Moderation APIs and webhooks: manage participants server-side and react to room events
• Egress and Ingress: record or multi-stream rooms and export tracks; ingest RTMP, WHIP, HLS, or OBS Studio streams
• Agents: build real-time multimodal AI participants that join rooms as backend peers
• Telephony: connect voice agents to inbound and outbound phone calls
• CLI: create tokens, call server APIs, and generate load-test traffic
• Deployment options: self-host as a binary, Docker, or Kubernetes, or run on LiveKit Cloud