| Real-time Voice | Sub-200ms latency with WebRTC. Continuous speech recognition, natural turn-taking, echo cancellation. |
| RAG | Upload documents for knowledge-grounded responses. Hybrid search (vector + BM25) with configurable similarity. |
| Conversation Memory | Persistent history across sessions. Automatic session tracking, per-user isolation. |
| MCP Tools | Connect external tools via Model Context Protocol. Function calling, custom tool definitions. |
| Multi-provider | Flexible LLM, STT, TTS configuration. Swap providers without code changes. Per-agent preset selection. |
| Webhook Events | Async event emission for session lifecycle, transcription, and response events. |
| Admin Panel | Web UI for managing agents, presets, and viewing conversations. |
| Error Tracking | Error logging, classification, and resolution tracking per session. |
| Self-hostable | Run on your infrastructure. PostgreSQL backend, bring your own LiveKit. |
| Text Chat API | Test agents via text without voice. Same agent config, direct text in/out. |