Voice & Telephony

Real-time voice input and output. Four channels cover the matrix: inbound calls, local microphone wake, outbound speech synthesis, and SIP-grade real-time conversation.

ClawdTalk (real-time SIP)

Full-duplex SIP voice powered by Telnyx. The agent talks over a real phone call (inbound or outbound). Supports barge-in, mid-turn tool use, and regional number provisioning.

zeroclaw config set channels.clawdtalk.<alias>.allowed_destinations <value>

export ZEROCLAW_channels__clawdtalk__<alias>__allowed_destinations=

zeroclaw config set channels.clawdtalk.<alias>.api_key <value>

export ZEROCLAW_channels__clawdtalk__<alias>__api_key=

zeroclaw config set channels.clawdtalk.<alias>.connection_id <value>

export ZEROCLAW_channels__clawdtalk__<alias>__connection_id=

zeroclaw config set channels.clawdtalk.<alias>.excluded_tools <value>

export ZEROCLAW_channels__clawdtalk__<alias>__excluded_tools=

zeroclaw config set channels.clawdtalk.<alias>.from_number <value>

export ZEROCLAW_channels__clawdtalk__<alias>__from_number=

zeroclaw config set channels.clawdtalk.<alias>.webhook_secret <value>

export ZEROCLAW_channels__clawdtalk__<alias>__webhook_secret=

api_key (Telnyx) and webhook_secret are secrets:

channels.clawdtalk.<alias>.api_key is a secret. Stored encrypted, never in plain config.toml. Set it through one of these, which encrypt on write:

zeroclaw config set channels.clawdtalk.<alias>.api_key    # prompts for masked input, stores encrypted

Pair with: a telnyx model provider for the brain and ensure your Telnyx account has a SIP connection with the correct webhook URL pointed at the ZeroClaw gateway.

Voice Call (Twilio / Telnyx / Plivo)

Traditional carrier voice: the agent picks up, transcribes the caller, replies with TTS. Higher latency than ClawdTalk but works with any regular phone number and doesn’t require SIP trunk provisioning. Outbound calls hit from_number and require operator approval when require_outbound_approval is on.

zeroclaw config set channels.voice_call.<alias>.account_id <value>

export ZEROCLAW_channels__voice_call__<alias>__account_id=

zeroclaw config set channels.voice_call.<alias>.auth_token <value>

export ZEROCLAW_channels__voice_call__<alias>__auth_token=

zeroclaw config set channels.voice_call.<alias>.excluded_tools <value>

export ZEROCLAW_channels__voice_call__<alias>__excluded_tools=

zeroclaw config set channels.voice_call.<alias>.from_number <value>

export ZEROCLAW_channels__voice_call__<alias>__from_number=

zeroclaw config set channels.voice_call.<alias>.max_call_duration_secs <value>

export ZEROCLAW_channels__voice_call__<alias>__max_call_duration_secs=

zeroclaw config set channels.voice_call.<alias>.model_provider <value>

export ZEROCLAW_channels__voice_call__<alias>__model_provider=

zeroclaw config set channels.voice_call.<alias>.require_outbound_approval <value>

export ZEROCLAW_channels__voice_call__<alias>__require_outbound_approval=

zeroclaw config set channels.voice_call.<alias>.transcription_logging <value>

export ZEROCLAW_channels__voice_call__<alias>__transcription_logging=

zeroclaw config set channels.voice_call.<alias>.tts_voice <value>

export ZEROCLAW_channels__voice_call__<alias>__tts_voice=

zeroclaw config set channels.voice_call.<alias>.webhook_base_url <value>

export ZEROCLAW_channels__voice_call__<alias>__webhook_base_url=

zeroclaw config set channels.voice_call.<alias>.webhook_port <value>

export ZEROCLAW_channels__voice_call__<alias>__webhook_port=

Voice Wake (local wake-word)

Runs locally, listens on the mic, triggers agent interaction when it hears the wake phrase. Useful for:

• Physical voice assistants on SBCs
• Desktop “hotword → ask” workflows
• Always-listening home-automation agents

The agent doesn’t send audio anywhere; wake detection is local. Only post-wake speech is captured and (separately) transcribed before reaching the LLM.

zeroclaw config set channels.voice_wake.<alias>.energy_threshold <value>

export ZEROCLAW_channels__voice_wake__<alias>__energy_threshold=

zeroclaw config set channels.voice_wake.<alias>.excluded_tools <value>

export ZEROCLAW_channels__voice_wake__<alias>__excluded_tools=

zeroclaw config set channels.voice_wake.<alias>.max_capture_secs <value>

export ZEROCLAW_channels__voice_wake__<alias>__max_capture_secs=

zeroclaw config set channels.voice_wake.<alias>.silence_timeout_ms <value>

export ZEROCLAW_channels__voice_wake__<alias>__silence_timeout_ms=

zeroclaw config set channels.voice_wake.<alias>.wake_word <value>

export ZEROCLAW_channels__voice_wake__<alias>__wake_word=

Build flag: Voice Wake is gated by the voice-wake cargo feature on zeroclaw-channels. Build with --features voice-wake to include it.

TTS (outbound speech synthesis)

TTS is an output service channels call into, not its own inbound channel. Global defaults live under tts. TTS provider instances are configured under providers.tts.<type>.<alias> (OpenAI, ElevenLabs, Google, Edge, Piper) and selected per agent via the agent’s tts_provider. See Model Providers for the provider entries and per-agent wiring. Provider API keys are secrets; set them through the gateway, zerocode, or zeroclaw config set, never in plaintext.

Latency budget

Speech feels real-time below ~500 ms end-to-end. Practical budgets:

ClawdTalk shortcuts several of these by keeping the audio stream live; regular voice_call incurs STT + LLM + TTS sequentially.

STT

Speech-to-text is configured separately from the voice channels; see the [transcription] config in the Config reference. Voice channels invoke whichever transcription provider is active when they need to turn audio into text.

Hardware notes

For always-on voice on an SBC:

• USB mic: any UAC-compliant mic works. arecord -l to verify the OS sees it.
• Speaker: either USB audio out or the SBC’s onboard jack; pick the OS default device for the user the daemon runs as.
• Microphones with built-in AEC (acoustic echo cancellation) dramatically improve wake reliability when the speaker is nearby.

See Hardware → Android for Android-specific audio setup.