You are handing over account access. Here is exactly what happens to it.
Connecting your socials, CMS, email or CRM to something that acts on its own is a real decision. This page is the detail behind it — written plainly, because vague reassurance is worth nothing here.
We never ask for a password
Access is granted through each platform's own permission screen — the same flow you would use to add a new employee or connect any other tool. We receive a scoped token, not your login. You can revoke it from that platform at any time, without asking us.
Credentials are scoped per agent
Tokens are stored encrypted in a secrets manager. An agent receives only the credentials its own task requires — the agent writing your listings never holds your CRM access. Secrets are kept out of prompts unless a specific run explicitly needs them, so they are not sitting in model context by default.
Nothing ships without your approval
Anything that leaves your business — a post, an email, a price — waits in your approval queue first. You decide which categories of work need a signature and which can run freely. Approval is the control that matters most, because it means a mistake is caught before a customer sees it rather than after.
Hard spending caps
Every agent has a monthly ceiling. When it is reached the agent stops, automatically, and we get alerted. Runaway cost is structurally impossible rather than merely unlikely, and it is our money at risk, not yours — AI running costs are inside your plan.
Everything is logged
Every task, decision and tool call is recorded in an append-only activity log. You can always answer “why did it do that”, trace it back to the goal it came from, and roll it back.
Your company is isolated
Each client runs as a separate company with its own data, budgets and audit trail. Nothing is pooled across clients. Your data is not used to train models, and we choose providers that contractually commit not to train on data submitted through their business APIs.
A human is accountable
We build it, we watch it, and we are responsible for it. When something looks wrong a person is already dealing with it. For regulated or reputation-sensitive work, the human review add-on puts a person between the agents and anything that reaches your customers.
What we do not do
- We do not require model provider API keys — the AI costs are ours by default. If you choose to bring your own key, it is stored and scoped exactly like any other credential, and the same spend caps apply.
- We do not see your card details. Payments go through Lemon Squeezy as merchant of record.
- We do not track you on this website: no cookies, no analytics, no third-party scripts.
If you leave
The work your company produced is yours. We export it, hand it over, and revoke our access to your accounts — and you can revoke it yourself at any point without waiting for us. Working data is deleted within 30 days of cancellation, sooner if you ask.
We are early. We do not hold SOC 2 or ISO 27001, and claiming otherwise would be a lie you could check. What we can tell you is exactly how the system works, which is what this page is for. If your procurement process needs a formal certification today, we are not the right fit yet — tell us on the call and we will say so plainly.
More detail on data handling is in the privacy policy. Anything this page does not answer, ask us on the contact page.