Understanding LiveKit Cloud architecture for per-user agents

Getting Started
1

This question originally came up in our Slack community and the thread has been consolidated here for long-term reference.

I need a clearer explanation of how LiveKit Cloud works.

Our MVP is four Python processes per user: one real-time agent that creates the room and talks to the user, plus three data-only agents that join the room and guide the real-time agent.

We’ve been running this locally, and now we need to scale and integrate it into an application. What’s the recommended LiveKit Cloud approach for deploying and running these agents per user?

2

Agents are generally per-user. You run one or more of your agent processes and dispatch users and agents to a room. LiveKit Cloud will load-balance across the available agent processes you have running.

It should not be much different than what you’re doing locally, except your agent process runs in the cloud.

See these docs for more details: