Thanks — great question. Yes, you can use a LiveKit agent to simulate scenarios with another agent to evaluate performance. To do this you’ll: • Configure the test room’s input so the evaluation agent accepts input from

AGENT

-kind participants (agents by default only listen to standard participants/SIP callers). For example, set

RoomInputOptions

with a participant filter that allows agent participants. • Script the evaluation agent with the scenarios/personas and prompts you want it to run (including edge cases). • Have the evaluation agent trigger tool calls (for example a

grade_response

tool) or log metrics so responses can be recorded and scored automatically. • Run repeated sessions to collect statistical results and refine prompts or agent logic. If you want, tell me: • Which SDK/language you’re using (JS, TS, Python, etc.) • Whether you need a sample config or a starter script for the evaluator agent I can then provide a brief example tailored to your environment and goals. Sources: How to test your agent using another agent, with interactive elements