Pick your stack on the left, customize prompts in the middle, talk on the right. Live latency, real per-call cost, transcript.
Per-stage time-to-first-token / first-audio-byte, measured during the call. Empty until the first turn lands.
Placeholders like {agent_name}
found in your greeting / prefix / base prompt. Whatever you enter here
replaces them before the call is sent — so the same template works for
many calls.
Describe what the voice agent should do — its role, the goal of the call, key flow notes. The LLM expands this into a polished prompt; voice-call rules, language and persona are added automatically as the prefix.