Your balances, debts, renewals, settings, audit history, and most working finance context are kept locally as the primary record.
Security and trust, in plain English.
Catalyst Cash is designed so your core financial record lives on your device first. This page explains what that means operationally, what leaves the device, and what controls you have.
What stays on the phone, what leaves it, and why.
AI requests and Plaid-linked refreshes are networked operations. If you don't use them, those requests don't happen.
Passcodes, device identifiers, API-style secrets, and subscription-state details are excluded from manual backup files.
How AI traffic is handled.
The app does not expose raw provider keys in the client. AI traffic is routed through the Catalyst backend proxy.
Native math runs first
The app computes key finance logic locally before the AI layer is used for explanation and structured decision support.
Context is minimized
Requests keep the card, merchant, and money context needed for useful answers while leaving out private account identifiers and access details.
No raw-payload retention
The product position is that raw financial payloads are not retained on Catalyst servers after request routing.
Provider training boundary
Current production AI runs through OpenAI's API. OpenAI states API data is not used to train models by default unless a customer opts in; Catalyst does not opt in to training on your API data.
Where Plaid fits into the security model.
Plaid is optional. If you use it, it improves freshness. If you avoid it, the app still works with manual entry.
Your banking credentials are entered into Plaid's interface, not directly into Catalyst Cash.
Plaid access-token handling stays off-device on the backend. The app primarily keeps local connection metadata and synced results.
If you connect Plaid, Catalyst uses linked-account data to power app features you choose. We do not sell that data or use it for advertising profiles.
You can stay fully manual if you prefer tighter control over what data ever leaves the device. Plaid's own handling is governed by Plaid's End User Privacy Policy.
The security controls you can actually touch.
Lock the app with a passcode and supported biometric flows like Face ID so casual device access is not enough.
Chat persistence can be reduced or avoided, and saved history can be cleared when you want a shorter local footprint.
Encrypted exports, Recovery Vault, linked restore flows, and secure-device continuity exist for portability without turning the product into a cloud-first finance database.
What this page is, and what it isn't.
This page explains product design decisions. It is not a third-party certification, formal security audit, or legal substitute for the Privacy Policy and Terms of Service.
Read the legal policy after the plain-English version.
This page explains the operating model. The privacy page covers the legal handling language.