There are no hard restrictions on creating multiple agents or agent sessions in LiveKit. The platform is designed to support multi-agent scenarios—including running multiple agents per server and deploying agents across multiple nodes. The architecture supports multi-agent handoff and simultaneous agent operation, and the underlying infrastructure scales to handle large numbers of concurrent sessions and participants. However, there are important considerations for performance and reliability: • Each agent session consumes server and worker resources. Worker capacity and resource availability (CPU, memory, network) may impact how many sessions can be run efficiently at once. • For production environments, it's recommended to run multiple servers per agent name, avoid sharing agent servers with other apps, and implement load balancing and auto-scaling to handle higher loads efficiently. • If session creation is taking a long time, you may be hitting resource limits or need to scale out your infrastructure. LiveKit also offers agent load testing features to help you measure and optimize concurrent agent performance. If you need assistance investigating slow session creation in your specific deployment, details like your deployment mode, number of agents, and resource metrics would help further diagnose performance bottlenecks. Sources: Voice Assistants | livekit/livekit_composite | DeepWiki | LiveKit Agents | README.md | test/agent_test.go | Building voice agents | LiveKit Docs | examples/voice_agents/llamaindex-rag/data/raw_data.txt