Voice interface for self-hosted AI assistants. Your server, your providers, your rules.
Have OpenClaw? Just tell your agent:
Install the Clack voice relay skill from https://github.com/fbn3799/clack-skill and set it up
Choose speech-to-text and text-to-speech providers separately. Mix on-device with cloud. Save money without sacrificing quality.
💡 Example: Free on-device transcription + premium ElevenLabs voices = great quality, zero STT costs.
Real-time, low-latency, with everything you need for natural AI conversations.
Speak naturally and get spoken responses. Automatic silence detection, adjustable sensitivity, interrupt anytime.
Built-in voice aliases for ElevenLabs. Or use iOS system voices for free on-device speech.
Use free on-device STT and only pay for premium TTS — or go fully local for zero API costs. You decide.
History persists across calls. Pick up where you left off — up to 50 messages of context.
Mute and unmute with play/pause. Full remote command center integration for hands-free conversations.
On-device STT supports all Apple Speech languages. Choose your recognition language in settings.
Clack is private by design. No data leaves your infrastructure.
Voice audio goes to your server — and only to the providers you choose. The developer never sees or touches your data.
No analytics, no tracking, no usage data. The app stores settings locally and that's it.
Constant-time HMAC token verification. One-time pairing codes with 5-minute expiry and rate limiting.
Add your own domain for automatic SSL. Encrypted WebSocket connections protect all data in transit.
Server, protocol, and app — fully auditable. Verify the security claims yourself.
No sign-up, no email, no personal information. Pair with a code and start talking.
Self-hosted means you own the infrastructure. Here's how to set it up.
Already have OpenClaw? Just tell your agent:
Install the Clack voice relay skill from https://github.com/fbn3799/clack-skill and set it up
Your agent handles cloning, dependencies, and service setup automatically.
Or set it up manually:
Set up OpenClaw on any server, VPS, or machine with a public IP.
npm i -g openclaw
Clone the voice relay skill into your workspace and configure your speech provider API keys.
git clone https://github.com/fbn3799/clack-skill.git ~/.openclaw/workspace/skills/clack
The script creates a venv, installs dependencies, and configures a systemd service.
bash scripts/setup.sh --domain clack.yourdomain.com
Get Clack from the App Store. Enter your server address in Settings.
Ask your agent for a pairing code, enter it in the app, and start talking. That's it.
Your AI, your voice, your rules. No subscription, no middleman.