LiveKit Agents is an open-source framework for building real-time voice AI agents that can hear, see, speak, and act. It is the Python framework behind a large share of production voice assistants, phone agents, and multimodal support bots. Developed by LiveKit and hosted on GitHub, where it has earned close to 15,000 stars, it is updated continuously, with activity through late 2026.
The framework lets you assemble a voice agent from the components you prefer: mix and match speech-to-text, LLM, text-to-speech, and realtime API providers instead of being locked into one vendor. Agents run as programmable participants on servers and connect to users through LiveKit’s WebRTC ecosystem, which covers web, mobile, and native clients. A distinctive strength is telephony: through LiveKit’s SIP stack, the same agent that talks to a browser can answer or place real phone calls. Built-in job scheduling and dispatch APIs route end users to available agent workers, semantic turn detection reduces awkward interruptions, and native MCP support lets agents call external tools with a line of code. A built-in testing framework with LLM judges helps you verify agent behavior before deployment, and the whole stack, including the LiveKit media server, can be self-hosted.
The Agents framework is free and open source under the Apache 2.0 license, and you can self-host the entire stack. LiveKit also offers LiveKit Cloud, a managed platform with paid tiers for teams that prefer hosted media infrastructure, usage-based scaling, and support, making the ecosystem freemium.
LiveKit Agents is aimed at developers building serious voice products: AI receptionists, voice support, telehealth assistants, and in-app voice companions. If you need production-grade realtime infrastructure with the freedom to choose your own AI providers, it is one of the strongest open frameworks available.