Often more than you think. An assistant wired to a shared service account can reach everything that account can — at conversational speed. Latten maps the blast radius from your real traffic: what each actor reaches, who asked, and where access runs wider than anyone intended.
See it on a live graph — no signup →Before you approve more AI, you want to know what it can touch. Latten shows the reach of every actor, drawn from real traffic, not a diagram of intentions.
When many actors share one service account, each inherits all of its reach. Seeing that on one graph is usually the moment the room goes quiet.
Latten is observe-only. It illuminates the reach; what you bound is your call, made with the full picture instead of a guess.
It is everything your AI can reach through the access it runs on — the real scope of what one prompt could touch.
No. It observes real traffic read-only and never sits in the request path.
No raw values leave the boundary — reach is mapped from types and counts, not contents.