Powerful AI agents need real controls, not blind trust. Here is how we protect your data today — and, just as importantly, what we are still building and what we do not claim.
Every Claw runs on a dedicated server, not a shared pool of users. Your agents, files and credentials are isolated from every other customer — there is no shared runtime that could leak one customer's context into another's.
The hub uses single sign-on. Access to a Claw's tools is scoped to the account that owns it — a session minted for one Claw cannot reach another's data.
Credentials and API keys are stored encrypted at rest in our vault and injected only where a task needs them — not published in page source or handed out in chat.
Each Claw can have its own email address on our own mail servers — separate from your personal inbox, so an agent never needs your password to send or receive mail.
Claws are backed up daily with a documented restore path, so a bad update is a restore, not a rebuild.
Security is never "done", so we publish the work in progress rather than implying it already exists:
We do not claim any system is unhackable or zero-risk — no honest provider can. An AI agent with real access is powerful, and powerful tools require care. Our approach is defence in depth, least privilege, and telling you plainly when we block something or when a control is still being built.
Found a problem? Tell us through our contact form and we will respond.
Claw Way is the managed-service brand; SQS Hub is the operating platform, operated by SQS. This page states our public position and is updated as controls ship.