How it works
Dial. Talk. Done.
Here's exactly what happens between you picking up the phone and hanging up — every system your voice touches, and what we keep.
The call path
From any phone — your call goes through your normal phone carrier.
A licensed SIP trunk routes the call into our voice infrastructure. Audio is encrypted in transit (SRTP / TLS).
Our system recognizes the caller (by phone number, then by voice if enrolled) and loads that person's private profile and history.
Your spoken words are converted to text in real time. We send only your audio — never your account profile — to our transcription layer.
Our reasoning layer decides whether to call a tool (look up zmanim, place an order) or just respond. Conversations are not used to train AI models.
A voice model speaks the response back. The whole loop typically takes under 700 milliseconds.
No app, no internet on your side — just the regular phone line.
What we do with your voice
- In transit: all audio is encrypted (TLS / SRTP).
- At rest: short-term retention (transcripts and brief summaries) lets your assistant remember context across calls. Specifics are in our Privacy Policy.
- Training: we do not use your voice or transcripts to train AI models.
- Selling: we never sell your conversations. Per CCPA, you can also opt out of sharing.
- Third parties: we use a small set of vetted infrastructure providers under signed DPAs. See the Security page.
What runs on your side
Nothing. There's no app to install, no account to set up online, no firmware. Any phone with dial tone works — kosher phones, flip phones, landlines, smartphones.