One agent, many languages

Pick a multilingual speech-to-text model and a voice that supports your target languages. The agent detects the caller's language and answers in it. Transcripts keep the original text plus an English translation.