Distributed multi-room audio routing. A software AV receiver that connects speaker nodes across a network, routes audio sources to any combination of rooms, and captures microphone audio for speech recognition.
Copyright (c) 2026 Skylark Software LLC. All rights reserved.
- Routes audio to any room — TTS, music, system audio, or any WebSocket source can be directed to one room, a group of rooms, or broadcast everywhere
- Per-room volume and mute — each node has independent volume and mute control from the dashboard or tray
- TTS switching — one-click to direct speech to a specific room, or broadcast to all
- STT microphone routing — select which node's microphone feeds your speech-to-text service
- Web dashboard — visual routing matrix, live node status, and audio source monitoring
- System tray — desktop nodes show connection status, mic toggle, and mute in the taskbar
- AI control via MCP — Claude and other MCP-compatible models can control routing, volume, and STT source
Audio Sources Skywire Server Speaker Nodes
───────────── ────────────── ─────────────
TTS engine ──────────────► routing matrix ◄────────► Living Room 🔊🎤
Music/MPD ──────────────► web dashboard ◄────────► Kitchen 🔊🎤
System audio ─────────────► REST API ◄────────► Bedroom 🔊🎤
MCP server ◄────────► Office 🔊🎤
STT engine ◄───────────── (mic audio)
The server runs centrally (Docker or standalone). Nodes are lightweight processes running on any machine with speakers — desktop machines use the tray app, headless machines use the node service.
Docker (recommended):
git clone https://github.com/Skylark-Software/Skywire.git
cd Skywire
docker compose up -dOr standalone:
chmod +x bin/skywire-server
./bin/skywire-serverThe dashboard is available at http://your-server:3030.
On each machine with speakers:
chmod +x bin/skywire-node
./bin/skywire-node --server ws://your-server:3030/audio --node-id kitchenOr install as a service (see Deployment below).
For desktop machines with a taskbar:
chmod +x bin/skywire-tray
./bin/skywire-trayConfigure the server URL in the settings dialog (right-click the tray icon).
Central routing engine. Handles all WebSocket connections, the audio routing matrix, and serves the web dashboard.
- Dashboard:
http://your-server:3030 - Audio WebSocket:
ws://your-server:3030/audio - Source WebSocket:
ws://your-server:3030/source - STT WebSocket:
ws://your-server:3030/stt
For desktop machines. Sits in the taskbar showing connection state — click to connect/disconnect, toggle mute, toggle microphone, or open settings.
 |
 |
| Right-click menu | Settings dialog |
- Blue icon = connected
- Gray icon = disconnected
- Red icon = error
- Mic checkbox enables microphone capture for STT
For always-on endpoints (servers, Raspberry Pi, etc.). Runs as a systemd user service with auto-reconnect.
./bin/skywire-node --server ws://your-server:3030/audio --node-id kitchenExposes audio routing controls to AI models via the Model Context Protocol.
{
"mcpServers": {
"skywire": {
"command": "./bin/skywire-mcp",
"args": ["--skywire-url", "http://your-server:3030"]
}
}
}Available tools: list_nodes, list_sources, get/set_routing, set_volume, set_mute, get/set_stt_source, get_status.
# Debian/Ubuntu
sudo apt install libxdo3 libasound2 libgtk-3-0
# Fedora
sudo dnf install libxdo alsa-lib gtk3# Debian/Ubuntu
sudo apt install libasound2
# Fedora
sudo dnf install alsa-libNo additional dependencies when running via Docker. Standalone requires only glibc 2.35+.
Pre-built binaries require glibc 2.35+:
| Distribution | Version | Status |
|---|---|---|
| Ubuntu | 22.04+ | ✅ Supported |
| Debian | 12 (Bookworm)+ | ✅ Supported |
| Fedora | 36+ | ✅ Supported |
| RHEL / Rocky / Alma | 9+ | ✅ Supported |
| Arch Linux | Rolling | ✅ Supported |
- x86_64 (amd64): Pre-built binaries provided
- aarch64 (arm64): Build from source required (Raspberry Pi 4/5)
Skywire uses ALSA directly on Linux:
- PulseAudio systems: Works via ALSA compatibility layer
- PipeWire systems: Works via ALSA compatibility layer
- Headless servers: Requires ALSA with a dummy or loopback device
Each node reads ~/.config/skywire/node.json:
{
"server": "ws://your-server:3030/audio",
"node_id": "kitchen",
"location": "Kitchen",
"sample_rate": 16000,
"volume": 100,
"muted": false,
"mic_enabled": true,
"mic_device": "pulse",
"auto_connect": true
}| Field | Description |
|---|---|
server |
Skywire server WebSocket URL |
node_id |
Unique identifier shown in dashboard |
location |
Human-readable room label |
sample_rate |
Audio sample rate (16000 recommended for STT) |
mic_enabled |
Capture and stream microphone audio |
mic_device |
Mic device name (pulse, default, or specific device) |
auto_connect |
Connect automatically on startup |
docker compose up -dThe server runs on port 3030 with host networking.
handle_path /apps/skywire* {
reverse_proxy localhost:3030
}
When behind a reverse proxy, nodes connect via:
ws://your-domain/apps/skywire/audio
[Unit]
Description=Skywire Audio Node
After=network-online.target sound.target
[Service]
ExecStart=/usr/local/bin/skywire-node
Restart=always
RestartSec=5
WatchdogSec=60
[Install]
WantedBy=default.target# Install binary
sudo cp bin/skywire-node /usr/local/bin/
# Enable service
systemctl --user enable --now skywire-node~/.config/autostart/skywire-tray.desktop:
[Desktop Entry]
Type=Application
Name=Skywire Tray
Exec=/usr/local/bin/skywire-tray
X-KDE-AutostartEnabled=trueSources connect via WebSocket to /source and send raw PCM or base64-wrapped JSON:
{ "type": "audio", "data": "<base64-pcm>", "targets": ["kitchen", "bedroom"] }STT engines connect to /stt and receive raw PCM from whichever nodes are selected as the active microphone source in the dashboard.
Copyright (c) 2026 Skylark Software LLC. All rights reserved.
This software is the proprietary property of Skylark Software LLC. Unauthorized copying, distribution, or modification is strictly prohibited. See LICENSE.
