Security

Coworkkit puts an AI co-worker inside your product, where it talks to your users and operates your interface. That only works if it is safe by construction. Here is how it is built, and where to reach us.

It runs as the signed-in user, under their permissions

The co-worker is not a separate account with its own powers. It acts inside your app as the user who is already signed in, through the actions and surfaces you choose to expose to it. It can do what that user can do, and nothing more. There is no elevated, behind-the-scenes access to your backend.

Your key never reaches the browser

The integration is front-end, but your Coworkkit key is not. It is exchanged through one server-side token route that you own, so the secret stays on your server and the browser only ever holds a short-lived session token. You control who gets a session, and when.

Consent, disclosure, and human hand-off are built in

Users are told they are working with an AI, actions that matter can ask for confirmation before they happen, and a session can hand off to a human. These are part of the runtime, not something you have to assemble.

Where your data goes

Sessions are hosted in the EU. Coworkkit uses a small set of sub-processors for the voice stack (speech-to-text, the language model, and real-time transport). The current sub-processor list, what personal data we process, why, and how long we keep it, all live in one place we keep current:

We are a young company and we would rather tell you exactly how the system is put together than wave a certificate at you. If your review needs something specific (a data-flow diagram, a sub-processor detail, a DPA), ask and we will walk you through it.

Reporting a security concern

Email hello@coworkkit.ai with "security" in the subject and we will get back to you quickly. If you believe you have found a vulnerability, please give us a chance to fix it before disclosing it publicly.