I'd be happy to help you implement a tool that lets the agent skip its turn and wait for the user to speak. You can achieve this by using manual turn control—just set

turn_detection="manual"

in the

AgentSession

configuration. This way, automatic turn detection is turned off, and you get to decide exactly when the agent listens and processes user input. Here’s a simple way to do it: • Use

session.clearUserTurn()

to clear any previous user input and reset the turn state, so the agent patiently waits for new input. • When you’re ready for the agent to respond, call

session.commitUserTurn()

to process the current user input. • If needed, you can also use

session.interrupt()

to stop the agent from speaking before waiting for the user. For example, you might create an RPC method like

start_turn

that your frontend can call to tell the agent to wait for the user:

ctx.room.localParticipant?.registerRpcMethod('start_turn', async () => {
  session.interrupt();        // Stop agent from speaking
  session.clearUserTurn();    // Reset user input state
  return 'ok';
});

This method gently instructs the agent to "skip" its current turn and wait as long as needed until

commitUserTurn()

is triggered. If you'd like to see this in action, there’s a complete example in the Push-to-Talk Agent. Feel free to explore it, and don’t hesitate to reach out if you have any questions along the way! Sources: Turn detection and interruptions | LiveKit Docs | examples/src/push_to_talk.ts