Autonomous Operations
A real adversary does not follow a static script. Kestrel reads what each host allows and picks the next sensible move, while you keep full control the entire way.
Autonomous, but never out of your hands
A good adversary does not follow a static script. Kestrel reflects that: its decision engine reads what it has learned from each host and picks the next sensible move, adapting the chain in real time.
You stay in control the whole way — run fully autonomous for breadth, switch to manual for a delicate step, pause to inspect state, or replay a run exactly as it happened. Everything is exposed through an open REST API.
- Adaptive decision engine. Chooses the next ability based on facts collected during the operation, not a fixed playbook.
- Manual or autonomous. Toggle between hands-off execution and step-by-step manual control at any point.
- Fully API-driven. Every operation, agent, and result is available over a documented REST API.
The decision engine, one cycle.
A real sequence that repeats until the objective is met or the operation is stopped.
Observe
Collect facts from each agent — users, hosts, privileges, reachable services.
Decide
Score candidate abilities against the facts and the profile's objective.
Act
Task the chosen ability to the right agent and execute it.
Learn
Feed the result back into the fact store, then repeat from the top.
Watch it think.
The operation view shows abilities executing, facts accumulating, and the chain branching in real time. Pause any time to inspect, or let it run to completion.
Where to go next
Launch an autonomous operation.
Set an objective, pick a profile, and let the decision engine work — with your hand on the controls the entire time.