Privacy

agents.host is made by AIUI Inc. (“we”, “us”), a company based in Quebec, Canada. This policy covers agents.host: the website, the dashboard, the assistants you build, the chat box you put on your site, the email addresses your assistants answer at, and our API.

There are two different relationships here, and they matter:

You, the account holder

You sign in, build assistants and run a workspace. We are responsible for the personal information in your account, and this policy says what we do with it.

Someone talking to an assistant

If you are chatting with an assistant that a business built on agents.host, that business decides what the assistant is for and what happens to the conversation. We hold it on their behalf. Ask them first; see “If you are talking to someone else’s assistant” below.

Our person in charge of the protection of personal information can be reached at support@agents.host. Write to that address for anything in this policy.

What you give us:

Your account

Your name, email address, profile picture and language preference. These come from Google or Microsoft when you sign in with them, or you give us your email address when you sign in by link.

Your workspace

Its name, logo, website, tagline and brand colours, and the email addresses of people you invite to it.

Your assistants

The name, the instructions you write, the tools and connections you turn on, the model settings, and the address it answers at.

Conversations

Every message sent to or from your assistants, on every channel — the dashboard, the chat box on your site, email, our API — along with any files attached to them.

Data lists

The files and records you upload for an assistant to read.

What you send us

Connection requests you post, and anything you write to support.

What we record as you use it:

Usage

Which model answered, how many tokens it read and wrote, which tools it ran, and what it cost. This is what your bill and the usage page are built from.

Sessions

When you signed in, from what IP address, and which browser you used. This is how we keep you signed in and spot an account being abused.

Memory

Short notes an assistant keeps about a conversation so it can pick up where it left off. You can read, edit and delete every one of them in Settings.

What we hold for you but never show:

Connection tokens

When you connect Google, Microsoft or an MCP server, we store the access token that connection issued. It is used only to make the calls your assistant makes. It is never shown to you, to your assistant, or to anyone else, and it is never sent to a model.

We do not collect payment card numbers. Stripe handles payment and we never see the card.

We do not run analytics, advertising trackers, or any third-party script that watches what you do on the site.

We use personal information to do the things you asked us to do:

• Run the service: deliver messages, run your assistants, keep the conversation history you can read back.
• Bill you, and show you what you spent before we do.
• Keep accounts secure and stop abuse of the platform.
• Answer you when you write to support.
• Meet obligations the law puts on us, like keeping tax records.

We rely on the consent you give when you create an account and again each time you turn on a connection, and on what we need to hold up our end of the agreement with you. You can withdraw consent at any time — see “What you can ask us for”.

We do not sell personal information. We do not share it for advertising. We do not build profiles of you to sell to anyone.

An assistant answers by sending the conversation to a language model. We reach models through OpenRouter, which passes the request to whichever provider runs the model you picked — Anthropic, OpenAI, Google and others. Which one is on the assistant’s page, and you choose it.

What goes to the model: the instructions you wrote, the conversation so far, and any file or data list content the assistant is allowed to read for that answer.

What does not go to the model: your connection tokens, your billing details, and other people’s workspaces.

We use these providers on terms that say the content is not used to train their models. We cannot audit their systems, and we tell you that plainly rather than promise something we cannot check.

Nothing is connected until you connect it. When you connect a Google or Microsoft account, you choose at that moment what the connection is allowed to reach, and your assistant can only use the tools you turned on.

You turn on one capability at a time — mail, calendar or files — and we ask for only the permissions those tools need:

Google mail (gmail.modify)

Read, search, draft, label and send mail in the connected mailbox, so an assistant can answer questions about your mail and reply on your behalf. It cannot permanently delete mail.

Google calendar (calendar)

Read and write events on the connected calendars, so an assistant can check availability, book, reschedule and cancel.

Google files (drive.readonly)

Read files in the connected Drive, and their names and metadata. Read-only: an assistant can find and read a document, and can never change or delete one.

Microsoft mail (Mail.ReadWrite, Mail.Send)

Read, search, draft and send mail in the connected mailbox, for the same reasons.

Microsoft calendar (Calendars.ReadWrite)

Read and write events on the connected calendars.

Microsoft files (Files.Read.All)

Read files the connected account can reach. Read-only, like Drive.

Signing in asks for less: your name, email address and profile picture, so we can create your account and show you who is signed in.

You can disconnect any connection from the Connections page. We delete the token when you do, and the assistant loses access immediately. What it already read stays in the conversation history until you delete the conversation.

This section covers what happens to data we get from Google when you connect a Google account. It adds to the rest of this policy; nothing here replaces it.

What we access is exactly the list above — Gmail through gmail.modify, Google Calendar through calendar, Google Drive through drive.readonly, and your basic profile when you sign in. We request a scope only when you turn on the capability that needs it, and never ask for one to hold in reserve.

How we use it

Only to run the assistant you configured, at the moment it answers a message in your workspace, through the tools you turned on. That is the entire purpose. It is not used for anything else.

How we store it

Google content is not copied into a library of its own. What an assistant read or wrote appears in that conversation's transcript so you can check what it did, and it lives under the retention rules in “How long we keep it”. The access token is stored so the connection keeps working, is never shown to you, your assistant or anyone else, and is deleted when you disconnect.

Who we share it with

The model that writes the answer, which receives only what that answer needs — see “What happens when your assistant answers” — and the infrastructure providers listed under “Who else touches it”, which store it for us under contract. Nobody else.

What we never do

We do not sell Google user data, use it for advertising of any kind, transfer it to data brokers or information resellers, use it to decide anyone's creditworthiness, or build a database out of it.

Human access

Nobody at AIUI Inc. reads your Google data. We would look at a conversation only if you explicitly ask us to — a support request you open — or where we must to investigate a security incident, or where the law requires it.

Deleting it

Disconnect the account and the token is gone immediately. Delete a conversation and the Google content quoted in it goes with it. Close your account and it is all deleted within 90 days.

On AI and machine learning: Google Workspace APIs are not used to develop, improve, or train non-personalized AI and/or ML models. Google user data reaches a language model only at the moment your assistant answers you, only to produce that answer for you, and only to deliver the feature you turned on. The providers who run those models are under terms that forbid training on the content we send them. We do not train models of our own on anything.

agents.host's use and transfer of information received from Google APIs to any other app will adhere to the Google API Services User Data Policy, including the Limited Use requirements.

Google API Services User Data Policy

You can disconnect a Google account from the Connections page in agents.host at any time. You can also revoke our access from your Google Account's third-party connections page, at https://myaccount.google.com/connections. Either one stops the assistant reaching your Google data.

The service is built on other companies’ infrastructure. Each of these handles some of your information under contract, and only to do the job we hired them for:

Cloudflare

Runs the service, stores uploaded files, and carries email in and out.

PlanetScale

Runs the database where accounts, assistants, conversations and usage records live.

OpenRouter and the model providers

Answer your assistant’s messages, as described above.

Stripe

Takes payment and holds the billing relationship.

Google and Microsoft

Sign you in, and provide the connections you choose to turn on.

Beyond those, we disclose personal information only when the law requires it — a court order, a warrant, or another valid legal demand — or where someone’s safety is at risk. If we are ever compelled to hand over your data, we will tell you unless we are legally barred from doing so.

Your information is stored and processed on servers run by the companies above, in the United States and in other countries where they operate. That means it may be handled outside Quebec and outside Canada, and may be reachable by the authorities of those countries under their own laws.

Before we add a provider that holds personal information outside Quebec, we assess whether it gives that information adequate protection, as Law 25 requires.

We keep information for as long as it is doing the job you gave it, and then we get rid of it:

Conversations, messages and files

Until you delete them. If you close your account, we delete them within 90 days.

Assistants and workspace settings

Until you delete them, or 90 days after the account closes.

Usage and billing records

Seven years, because tax law requires it. These say what was spent, not what was said.

Sessions

Thirty days after the session expires.

Connection tokens

Until you disconnect the connection, or the account closes.

Backups run on their own cycle and clear within 35 days, so deleted data can survive that long in a backup before it is gone for good.

Under Quebec law and Canadian federal law, you can ask us to:

• Show you the personal information we hold about you, and tell you where we got it.
• Give you a copy in a structured, commonly used format that you can take elsewhere.
• Correct anything that is wrong or incomplete.
• Delete it, where we are not required to keep it.
• Stop relying on your consent, and stop the processing that depended on it.
• Explain any decision made about you purely by automated means.

Much of this is already in the app: read every conversation under Conversations, see every charge under Usage, read and delete an assistant’s memory under Settings, and disconnect any connection under Connections.

For anything else, write to support@agents.host. We will answer within 30 days. If we say no, we will tell you why and how to challenge it.

If you are not satisfied with how we handled your request, you can complain to the Commission d’accès à l’information du Québec, or to the Office of the Privacy Commissioner of Canada.

Assistants generate their answers automatically — that is the whole product. We do not use them to make decisions about you that have a legal effect or that significantly affect you.

If a business has built an assistant that does make such decisions about you, that business is responsible for it, and it owes you an explanation. We will pass your request on to them.

Traffic to and from agents.host is encrypted in transit. Data at rest is encrypted by the providers that store it. Access to production systems is limited to the people who need it, and connection tokens are never rendered into a page or handed to a model.

No system is perfect, and we would rather say so than imply otherwise. If a breach happens and it creates a risk of serious injury, we will notify you and report it to the Commission d’accès à l’information du Québec, as Law 25 requires.

We use three, all of them necessary to make the site work:

The session cookie

Keeps you signed in.

locale

Remembers whether you want English or French. One year.

ah_visitor

Set only inside a chat box on someone’s website, so a visitor’s conversation survives a page reload.

Your theme choice is kept in your browser’s own storage and never sent to us. There are no advertising cookies, no cross-site tracking, and no third-party analytics on any page we serve.

agents.host is not built for children. Do not create an account if you are under 14. If we learn that we hold personal information about a child under 14 without the consent of a parent or guardian, we will delete it.

When you chat with an assistant in a chat box on a business’s website, or email one, that business wrote its instructions and decided what it can reach. They are responsible for what it does with your conversation and for telling you about it in their own privacy policy.

We store the conversation on their behalf, and we do not use it for anything of our own. If you write to us about it, we will forward your request to the business that runs the assistant and tell you we have done so.

When this policy changes, we update the date at the top of the page. If a change materially affects what we do with your information, we will email account holders at least 30 days before it takes effect.

Write to support@agents.host with anything at all — a question, a request to see or delete your information, or a complaint. It reaches the person in charge of the protection of personal information at AIUI Inc.