Security
Data flow
What data Katexs processes, where it goes, and what is stored.
During a call
Caller audio is streamed to the voice engine, transcribed by the transcriber you chose, sent as text to the model you chose, and the reply is synthesized by the voice you chose. Tool arguments are sent to the tool's endpoint.
Stored by Katexs
| Data | Stored | Notes |
|---|---|---|
| Call log, transcript, summary, analysis | Yes | Scoped to your workspace |
| Recording | Only if recording is on and HIPAA mode is off | Link on the call |
| Chat transcripts | Yes | On the chat session and contact timeline |
| Contacts and timeline | Yes | Built from calls, chats and bookings |
| Tool inputs and outputs | Redacted summaries | Phones and emails masked |
| Provider keys | Encrypted | Never returned in full |
Providers
Model, transcription and voice providers process conversation data to deliver the service, according to the providers you select for each agent. For the current list of subprocessors, or to discuss data residency, email support@katexs.com.
Last updated November 2026 · Found a problem? Email support@katexs.com
