Thanks — great question. I’ll answer the two parts you asked about. Short answer • Concurrent agent sessions are enforced at the project level, not per individual agent. • If you deploy two agents to LiveKit Cloud, the work will be distributed between them and sessions will be scheduled across those agents. Details that matter • The project-level limit means your plan’s concurrent-session cap (for example: Build = 5, Ship = 20, Scale = 600) applies to the total number of active agent sessions across all agents in that project. • The agent scheduler distributes jobs across available agents using a round‑robin / single‑assignment approach and considers effective load (user locality, CPU, memory, network) when placing sessions. • LiveKit Cloud uses stateful assignment: once a session is placed on an agent server it stays there for its duration to preserve state and timing; failover only occurs if the target server doesn’t accept the job in time. If you’re planning capacity • Deploying two agents will increase your available processing capacity because sessions will be spread between them, but your maximum concurrent sessions remain the project cap set by your plan. • To increase concurrent session limits, you would upgrade your plan or request higher limits via the dashboard. Would you like help estimating how many agents you should deploy for your expected peak concurrency and workload characteristics (codec, average session length, expected CPU/CPU+media load)? If so, tell me your target concurrent sessions and a bit about typical session resource use and I’ll help size it. Sources: Pricing | LiveKit | Deploy and scale agents on LiveKit Cloud | Self-hosted deployments | LiveKit Documentation | How much can LiveKit scale? | LiveKit