Pick an agent
Real-time voice conversation — speech recognition, tool-calling AI, streamed speech synthesis with barge-in. Text chat works too.
Mic requires Chrome/Edge. The browser captures 16 kHz PCM; the server runs VAD → Whisper STT → LLM tools → Piper TTS and streams audio back. Speak, or type below each agent card. Say something while the agent talks to barge in.