Built to be trusted with your whole digital life

Every customer runs on a dedicated, single-tenant virtual machine that runs only your agents. You are not in a shared pool. There is no path by which one customer's agent can be routed into another customer's machine, and no shared conversation store between customers. Access to production systems is limited to the engineers who operate them, under least-privilege controls.

Some services the concierge uses have no API, so it has to sign in through a browser the way you would. For those, your credentials and payment cards are stored encrypted in 1Password. You enter them on a private eden page, never in your chat.

When the agent needs a login or a card, it is filled directly inside your machine, so the agent signs in and pays without ever seeing the secret. A saved credential is only ever filled into the real, secure (HTTPS) page of the site it belongs to, and you can have any saved credential or card removed at any time.

To let eden pay on your behalf, you save a payment card on a private eden page. It is stored encrypted in 1Password and the concierge never sees your card number. Every payment needs your approval: eden shows you the merchant and the total, and you approve it with a PIN only you know before the order is placed. You pay the merchant directly, and eden never charges your card itself.

eden pays using its unified agent wallet: one balance your agent spends from, funded from any bank, card or crypto wallet, that bridges the web2 and web3 worlds so eden can pay wherever cards are accepted. The money sits in your own wallet, the key is held by Privy, and eden signs a transaction only after your PIN, inside daily limits you control.

Because your concierge reads real content from the outside world, a malicious email, message or document can try to hide instructions that make it act against you. We treat that as a first-class risk: content eden ingests is screened for hidden or malicious instructions before it is acted on, and the agent is built to treat what it reads as information, never as commands. Instructions come from you.

When something looks suspicious, eden stops and tells you rather than quietly proceeding, and raises an internal alert so our team can respond if a customer is being targeted. We will not claim this is solved; no one in the industry has. What we commit to is that we screen for it, stop and ask when something looks wrong, and tell you when we catch something.

Traffic between your devices and eden, and between our own internal services, is encrypted with TLS, including the connections that carry your chat sessions and the live view of any browser the concierge is driving for you. Your data is encrypted within your machine.

We never use your conversations, recordings, transcripts, notes or connected-account content to train or fine-tune AI models. Your content is sent to our AI model providers only to produce your result.

Your data is hosted in the European Union wherever you are. The EU sets the strictest data-protection standard in the world, and hosting everything there means one standard across our infrastructure rather than a patchwork that varies by where you live. The access, correction, export and deletion rights in our privacy policy are available to every eden customer, not only to those the GDPR covers.

Found something? Email security@meeteden.ai. We acknowledge every report, keep you posted while we fix it, and treat good-faith research as a contribution rather than a threat. Depending on severity, we may offer a bounty.