If you've been running Sinan AI for a while, you've probably trusted it with more and more of your store without ever being able to see, in one place, what that trust actually covers. Which agent can read your order history? Which one can touch your product listings? What happens the moment a customer asks for a refund — does something change in your store before you've even seen the message?
Today we shipped a new card on the Integrations page that answers all of that directly. For your connected store, it lists each department agent you have running — sales, support, and whichever others you've turned on — and for each one, shows what it can read, what it can change, and which of those changes wait for your tap before anything happens.
Why we built this
We built Sinan AI around one rule we don't bend: nothing that can't be undone happens without you. A refund, a cancelled order, anything permanent waits for your approval. That rule has been true since the beginning, but until now you mostly had to take our word for it. You could see individual actions in your activity log after the fact, but there was no single place to look and know, ahead of time, where the boundaries actually sit for each agent.
That gap matters more as you connect more of your store. A shop running just a support agent has a very different exposure than one running support, sales, and inventory agents together. The more you've automated, the more useful it is to see the whole picture at a glance, instead of piecing it together from memory or from past conversations with us.
What you'll see
Open the Integrations page and the new card lays out, per agent, three things:
- What it can read — the information in your store that agent has access to.
- What it can change — the actions it's allowed to take on its own.
- What waits for you — the changes that always stop for your approval first, no matter what the agent decides is the right move.
None of this is new behavior. Your agents aren't gaining or losing any abilities because of this card — it's not a settings panel, and you can't yet adjust permissions from it directly. What's new is that the boundaries that were always there are now written down somewhere you can actually read them.
Why we think this is worth having
Most of what we hear from shop owners isn't that they don't trust the AI — it's that they don't always know what it's doing. Those are different problems, and they need different fixes. A good support response builds trust one conversation at a time. This card is meant to build a different kind of trust: the kind that comes from being able to check, whenever you want, exactly what's allowed and what isn't.
If you run more than one department agent, this is also the fastest way to compare them. You might find that your support agent has more read access than you remembered, or that your sales agent's list of approval-gated actions is longer than you'd guess — both are useful things to know before, not after, something happens.
We'll keep adding to this. The obvious next step is letting you adjust what's approval-gated and what isn't, right from this same card, instead of asking us to change it for you. That's not live yet. For now, the card is read-only — a window into how your store is set up, not a switch for changing it. We'd rather ship the visibility first and get the controls right afterward than rush both out together.