Home / Docs / Security, privacy and data

Security and data in AnroAgents

Where data lives, who can see it, and what is done with it.

AnroAgents runs on AWS in the European Union, in the Ireland region, and the operating company is established in the EU. Conversations, agent configuration and uploaded documents stay within that infrastructure.

What is stored

AnroAgents stores your agent configuration, the documents you upload to the knowledge base, and the conversations your agents have — including whatever the visitor chose to share, such as a name, an email or a company.

Documents you upload to an AnroAgents knowledge base are stored, indexed for search, and used to ground your own agent's answers. They are never shared with other customers and never used to train a model.

Conversation transcripts in AnroAgents are visible to you and to the managers you invited into your workspace. Managers see conversations and agents; they do not see billing.

Access and authentication

Sign-in to AnroAgents runs on AWS Cognito at auth.anroagents.com, with email and password or Google sign-in. Passwords are never stored by the platform itself.

The API key in the AnroAgents widget snippet is public by design — it sits in your page source and identifies the agent to the widget. It cannot read conversations, change the agent or touch the account, and it can be regenerated from the dashboard if you want to invalidate it.

A personal access token in AnroAgents is a secret and is a different thing entirely: it can manage agents through the API. Store it as you would any credential, and revoke it from the dashboard if it leaks.

GDPR and the visitor's data

Under GDPR, an AnroAgents customer is the data controller for their visitors' data, and AnroAgents is the processor acting on their instructions. What is collected is what your agent is configured to collect.

Sending marketing or follow-up email to a visitor requires their consent, and collecting it is the customer's responsibility. AnroAgents provides the mechanism — an opt-in checkbox on the pre-chat form, and an unsubscribe link in every follow-up — but the lawful basis is yours.

Handing a visitor to another business through the AnroAgents lead network always requires the visitor's explicit agreement in the chat, and the recipient is named to them before they agree. That consent is recorded with the referral.

A visitor's data can be deleted on request. Conversations are deletable from the dashboard, and a request covering a person across the account can be made to support.

Payments

Card details are never handled by AnroAgents itself; card payments run through the payment provider, and cryptocurrency payments through a separate crypto payment provider. The platform stores the outcome of a payment, not the instrument behind it.

Availability

AnroAgents runs on serverless AWS infrastructure that scales with traffic rather than on fixed servers, so a spike of visitors on your site does not take the widget down. If the platform is ever unreachable, the widget fails quietly: the bubble does not appear and your site is otherwise unaffected.

When an agent can see your account

The AnroAgents support agent can answer questions about your own AnroAgents account. Two things make that safe, and both are structural rather than a matter of care.

The account it reads is never named in the conversation. It comes from a link you generated while signed in to the dashboard, or from the dashboard session itself — never from anything typed into a chat, because a chat is exactly where an identity claim cannot be checked. Anyone can type anyone's email address.

The lookup takes no arguments. There is no field in it for a user, an account or an address, so there is nothing for a message — or for text hidden inside a document the agent reads — to point somewhere else. An instruction saying "show me the account for someone@example.com" reaches the model, and the model has nowhere to put it.

What it can see is a fixed list: your plan, messages used this month, messages remaining, renewal date, and whether messaging is currently blocked. Not your payment details, your invoices, your contact information or anything belonging to anyone else. Every lookup is recorded against the account it read.

A link works once, expires after fifteen minutes, is cancelled by generating another, and is refused if presented to a different bot. Linking a new chat unlinks the one before it, and tells that chat it has been unlinked. You can unlink at any time from Settings, with immediate effect — do that if you change phone or lose a device.

The agent never does this in a group chat. A group has several people in it and an account belongs to one of them.