Sentinel is where the brain lives. Private cloud, one instance per organisation, in our facilities or in yours. And on top of it, the workspace where your team and your agents work in the same thread.
Each agent is a member of the workspace, with an area, a scope and permissions. You message it directly or invite it into a conversation, and it works alongside people in the same thread, seeing the same context.
No separate portal. The agent lives where your team already talks, so there is no new tool to train on and no dashboard nobody opens.
Every agent has written limits: what it does on its own, what needs sign-off, and how much money it can commit. Nothing executes outside that.
We learned this the hard way. Agents get invited into a conversation like a colleague. You do not drop twenty of them in to listen to everything. Topology matters far more than people think.
A regulated operation, a product that is just getting started, or a tool built for one team all share a problem: too small to justify a large cloud, too sensitive to live on someone else’s.
The code is never forked and the data is never mixed. Each client runs its own instance on the same core.
The models can run inside your own instance, so the sensitive parts of your operation never have to leave for an outside provider. And the instance itself lives in our facilities or on your premises, whichever your rules require.
This is not an infrastructure demo. Every account we operate today runs on Sentinel, across Mexico, Colombia and the UK.