Defend from within · Probe from without
Two VIs working the same wall from opposite sides. Vigil watches from inside your systems and blocks attacks as they form. Vendetta attacks from outside — safely, on your behalf — finding the holes first and handing you the patch.
Attackers don't schedule. Vigil and Vendetta run without pause — one hardening from the inside, one stress-testing from the outside — so the gap between "vulnerable" and "patched" shrinks toward zero.
Coverage
Both VIs run continuously against your live surface.
Validation
Vendetta stops at proof — it demonstrates, it never detonates.
Control
A one-switch stop and full audit trail on every run.
Output
Every finding ships with a documented, prioritized patch.
Given access to a system or architecture, Vigil builds a world model of it from within — every service, trust boundary, and data path — then works to keep attackers out before they get a foothold.
Scans from within to map the real attack surface.
Vigil enumerates your architecture from the inside: hosts, services, dependencies, identities, and the paths between them. It sees the misconfigurations and forgotten doors an outside scan never reaches.
That internal map is a world model — kept live as your infrastructure changes.
Learns normal, so it notices the moment something isn't.
By modelling ordinary behaviour — traffic, access patterns, process trees — Vigil spots the anomaly that signals an intrusion in progress, not just a signature it was handed.
Novel attacks look wrong before they're named. Vigil is built to feel that wrongness early.
Blocks proactively, within the guardrails you set.
When Vigil identifies an active threat it can act — isolate a host, revoke a token, tighten a rule — inside a policy you define. High-impact actions can require a human, or run autonomously where you've allowed it.
Speed where it's safe; a person in the loop where it counts.
Hardens continuously and reports what it changed.
Every vulnerability Vigil finds becomes a prioritized, documented fix, with a clear audit trail of what it did and why. Your posture improves a little every day instead of once a quarter.
Defence as a habit, not an event.
From the outside, with nothing but what a real adversary would have, Vendetta probes your perimeter, chains what it finds, and proves the risk — then stops, documents it, and leaves the patch on your desk before anyone hostile arrives.
Enumerates the perimeter like an attacker, automatically.
Vendetta starts where a stranger starts: your externally exposed surface. It discovers assets you forgot you had, fingerprints them, and prioritizes the ones worth pushing on.
Ownership is verified first — Vendetta only tests what's yours.
Chains exploits step by step, each move built on a proven last one.
Real breaches are chains, not single bugs. Vendetta reasons across steps — a leak here enabling a foothold there — the way a skilled human red-teamer would, at machine breadth.
Rate-limited and scoped, so testing never becomes the incident.
Stops at proof. It demonstrates the risk; it never detonates it.
Vendetta halts at a non-destructive proof-of-concept — enough to show the path is real, never enough to cause harm. A kill switch and complete audit trail govern every run.
You get certainty about the danger without living through it.
Documents everything and hands Vigil the fix.
Each finding becomes a replayable evidence chain mapped to the frameworks you report against, plus a concrete remediation. Vendetta finds it; Vigil helps close it; a free retest confirms it's gone.
Offence and defence, feeding each other on a loop.
Vendetta verifies ownership before it touches an asset. It tests what you prove is yours, nothing else.
Validation stops at proof-of-concept. No data is exfiltrated, corrupted, or left changed.
Every run is rate-limited, fully logged, and stoppable in one action, at any moment.
These are tools for the defender. We deploy them for organizations testing their own systems — not to enable attacks on anyone else.
Vigil and Vendetta are VIs: autonomous within strict scope, but never self-directed. They don't choose their own targets, can't act outside an authorized engagement, and keep a human accountable for every consequential decision.