Context
Current assets, facts, credentials, prior results, objectives, and the risk policy in force.
THE DARKO ENGINE
Structured attack knowledge. Adaptive tactical decisions.
Darko combines deterministic execution controls, evidence-driven state, reusable attack knowledge, and model-assisted reasoning to decide what an engagement should investigate next.
Book a DemoLive engagement
Watch Darko reason through an engagement.
Autopilot runs approved tests on its own; Guided pauses for your review before each action. Same engine, the level of control set by you.
DarkoAutonomous Pentester14:35:42
NTDS Extraction›192.168.10.12
Extracting NTDS from the domain controller.
Domain Compromise
Pwn3d!
DarkoAutonomous Pentester14:36:15
Trust Enumeration›essos.local
Trust enumeration from essos.local complete.
DarkoAutonomous Pentester14:37:07
AD Graph Collection›192.168.10.12
BloodHound collection for essos.local complete.
DarkoAutonomous Pentester14:38:19
Host Credential Harvest›192.168.10.10
Credential harvest from 192.168.10.10 — nothing found.
Autopilot — runs approved tests automaticallyGuided — waits for your review before acting
Evidence in · next move outHow Darko decides
New evidence changes what should happen next.
A credential, a validated access tuple, a blocked path, a new trust relationship, or a failed prerequisite can change candidate priority — adapting to the environment without granting free-form execution authority.
Candidates
Valid actions whose prerequisites and required arguments can actually be satisfied right now.
Decision summary
A concise rationale for the selected next action — not hidden reasoning presented as evidence.
What holds it together
Adaptive, but never unaccountable.
The model is part of the system, not the whole system.
Tools execute the work. Ingesters turn results into structured state. Control-plane logic handles known requirements and forced transitions. Darko uses the resulting evidence when a tactical choice is not purely deterministic.
Known technique, reusable structure.
Hundreds of structured attack families describe prerequisites, arguments, risk, expected output, and evidence. Chain logic connects repeatable multi-step paths without asking a model to invent the mechanics.
Exact catalog counts belong in dated release notes, because the catalog changes over time.
Claims still require evidence.
Model confidence is not proof. Customer-facing claims are grounded in parsed results, attributed artifacts, and verification contracts appropriate to the claim.
NEXT STEP
See how Darko would reason through your testing objective.
Book a focused walkthrough with our team.
Book a Demo