agents

LiveKit Agents is an open-source framework designed for building realtime, programmable, multi-modal voice agents that run on servers. It enables agents to see, hear, and understand with low latency. The framework offers flexible integrations for STT, LLM, TTS, and Realtime APIs, built-in job scheduling, semantic turn detection, and telephony (SIP) support. It also natively supports the Model Context Protocol (MCP) and provides a dedicated testing framework.

• Realtime Multi-modal Interaction (Voice, vision, and data exchange via RPC)
• Flexible Model Integrations (Mix & match STT, LLM, TTS, and Realtime APIs)
• Native MCP Support (Integrate Model Context Protocol tools with one line of code)
• Telephony & WebRTC Integration (Seamlessly make or receive calls via SIP)
• Built-in Testing Framework (Automate and judge agent behaviors for reliability)