Voice + SMS
The same agent you build today will soon answer the phone and reply to texts. Same Soul, same eval gate, three new transports.
Coming soon
Voice and SMS transports are on the post-launch roadmap. The infrastructure (Twilio integration for SMS, Vapi/Retell for voice) is partially in place — what's left is wiring the agent runtime's Soul/Brain/tools loop into the phone-call flow and adding voice- specific eval scenarios.
What it'll look like
Imagine you've built and published an HVAC chatbot today. With voice and SMS turned on:
- SMS. A customer texts your business number (forwarded through Twilio). The same agent picks up the conversation. It can book, reschedule, look up history, escalate.
- Voice. A customer calls the same number. The agent answers in a natural voice (your choice of provider — ElevenLabs, OpenAI Realtime, Vapi). Real-time, sub-500ms latency. Same booking tools, same eval-gated safety.
- Unified inbox. All three transports (web chat, SMS, voice) feed into the same Conversations view, threaded by contact. You read a customer's full history regardless of channel.
Why we're shipping this last
Voice and SMS are the highest-stakes channels — a hallucinated appointment over the phone is worse than one in chat because the customer trusts the voice more. We're holding the launch on voice/SMS until we've extended the eval suite with channel-specific scenarios (interruptions, accents, hold music, dropped calls, late-night SMS) and proven the runtime regenerates on critical fails before the customer hears the wrong answer.
Get notified when it ships
Subscribe to the GitHub releases feed — voice/SMS will land in a v1.4x release with full migration notes. Existing agents won't need changes; you'll just toggle the transport on per-agent.