What it does

Everything between the microphone and the reply.

Core capabilities, one router. Compose them per agent, pair clients securely, and observe which provider actually served the turn.

Wake word + barge-in

An openWakeWord listener runs always-on. Say a wake word mid-reply and the whole playback process group is signaled instantly to start the new turn — barge-in is wake-gated, so it works without echo cancellation. Custom “hey dante”-style models train from synthetic speech alone in about 30 minutes — no recordings needed.

Streaming TTS

Audio starts the moment the provider emits its first chunk via /v1/tts/stream — no waiting for full synthesis.

Swappable providers

STT, LLM, and TTS are all pluggable. Cloud APIs or fully local servers — any OpenAI-compatible endpoint drops straight in.

OpenRouter sign-in

Authorize OpenRouter from the local Settings UI with a one-time PKCE flow. The hub keeps the verifier server-side, claims the issued key once, and never puts it in a URL or browser history.

A2A routing, per-agent voices

Route turns to OpenClaw / Hermes peers over A2A. Define peers by IP or let recommended setup migrate a compatible OpenClaw 2026.7.1-2 config, scan private networks, and verify locally stored credentials against gateways that publish a public card but protect conversations — then save the working agent without secret copying.

Provider health

See process-lifetime attempts, errors, 429s, fallback rescues, last use, and supported quota headers in the console or /v1/providers/stats.

MCP tool calling

LLM agents call tools over MCP servers automatically — up to 5 rounds per turn, speaking JSON-RPC 2.0 over streamable HTTP with session continuity. Unreachable servers are skipped, not fatal.

Per-agent memory, screened before replay

Each agent remembers its last N turns, optionally persisted to disk. OWASP Agent Memory Guard screens stored history for prompt-injection patterns and redacts token-like secrets before a later LLM fallback can replay that context.

Self-hosted, local-first, client-safe

The always-on pieces stay on your hardware. Docker Compose, Windows and macOS desktop hubs, local-CA HTTPS for paired LAN clients, and revocable phone/watch credentials instead of administrator-key copying.

docker desktop preflight pairing console traces

Ready to route your own voice stack?