Now that computer use agents are at a functional level, does livekit support a computer use tool or handing off a task to a subagent in a vm?
Maybe I’m not following, but you’re asking about spawning a subagent to perform some long running task like deep research? Those kind of long-running operations would use an async task, and we have other workflows documented here. For a subagent in a VM, I think you would need to expose something through MCP and have the agent access it that way since, although we support agent handoff, those are still within the same AgentSession.
What use case were you looking to solve with subagents?