Security and your data
Every agent runs in its own isolated environment. It sees your data and nobody else's, and it can only reach systems you have connected yourself.
Where does my data live?
In the agent's own isolated environment: its memory and its files are kept apart from every other customer's, so one customer's data is never in another's reach. If you want nothing kept on our side at all, the agent's data can live entirely on your own server (available with a capacity tier): see keeping data on your own server.
What can the agent touch?
Exactly what you gave it: the channels you connected, the documents you shared with its address, the accounts whose access you handed over. Nothing else exists for it.
How do I take access back?
The same way you gave it. Unshare the folder in Google, remove the agent from the chat, revoke the key in the service. It takes effect immediately — there is nothing to ask us for.
How are secrets handled?
For passwords, tokens and login codes the agent issues a one-time secure page, and the secret goes straight into its protected storage instead of the chat history. If someone asks you to paste a password into a conversation, that is wrong — ask for the secure link instead.
What about payment data?
Payment is by card on the payment provider's page, by crypto transfer or by bank transfer against an invoice, so no card details pass through us.
Can the agent act without me?
It acts within the job you gave it. Paying, deleting, sending something on someone else's behalf, entering a password or card details, and installing or removing programs on your computer wait for your explicit yes. Anything else you want to see first, such as writing to a client or publishing, you put under a rule in plain words: "ask me first". Tell it once and it holds.