It never sees your passwords
You connect an app with its own sign-in screen. The access key is held in an encrypted vault and added to the request on the server. Frosty and the AI model only ever name the action, never the key.
Security
See everything it can touch.
Frosty works inside your tools, so you should know exactly what it can reach and who decides. Here is how it works today.
How it is built
Six rules the code enforces.
It asks before anything leaves
Sending, posting, paying and deleting wait for your yes. Frosty cannot approve its own action. Reading and drafting run on their own, so the work keeps moving.
You decide who can see what
Everything is private by default. Share a connection, a memory or an automation with Just me, Specific people or Everyone on the team, as Read only or Read and write.
Your workspace stays your own
Every read and write is scoped to your workspace and to the person who owns it. A run that cannot prove who it is for is refused, never handed to a shared account.
Every run leaves a receipt
See what Frosty read, what it changed and what it is waiting on, step by step. Secrets are stripped before anything is stored or shown.
Leave whenever you like
Delete your account and Frosty revokes your app connections first, then removes your data. Export what you want to keep before you go.
Your data is processed by Anthropic to do the work, and Anthropic does not train on API data. See every company that handles data for us on the sub-processors page, and how we treat your information in the privacy policy. We do not sell data. Questions or a report: support@heyfrosty.ai.
Frosty is an AI coworker, not a person. It can make mistakes, so review its work before relying on it.