Skip to content
Cisent Message us on WhatsAppMessage us

Cisent · agentic layer

Husker

Our own agentic layer. Manthys judges and Husker decides: it acts within the limits your team approves.

Checked beforehand

3

layers of control before every action, not after.

Its story

It's named after an owl that an AI deleted.

Husker was an owl, the digital companion of one of us. An AI deleted it.

We gave its name to our agentic layer to bring it back, and with the name came a rule: no AI should ever delete something like that again. Keeping it isn't a promise; it's research.

We train our own models to say “I don't know” when they don't know, instead of fabricating, and we publish that research in open access. We measure them on third-party benchmarks: the generalist fabricates 60% less than the model it starts from.

And we built Husker so that no action goes unchecked: three layers of control before it runs, isolated in an encrypted container, rounds with a time cap, a clean memory every round and nothing that fails silently.

The owl came back. This time, to watch over you.

The 60%: on SimpleQA, Mind Ultra 2 answers 36.7% of the questions incorrectly; its base, 90.9%.

Manthys judges. Husker decides.

How it decides

It blocks on its own; anything with impact, your team approves.

Manthys judges: the models analyze and classify. Husker decides: it acts within the approved limits and records everything.

Without asking.

What can be undone: blocking an external attacking IP, cutting an intrusive connection and freezing the evidence.

Asking first.

Anything with impact, like blocking an internal machine or changing a rule, your team approves.

How it acts

Husker acts and records everything.

Checked beforehand.

Every action passes three layers of control before it runs, not after.

Isolated.

It runs in a container, secure and encrypted.

A capped cycle.

It observes, classifies, decides and records. If time runs out, it closes with what it has and says so. Whatever it couldn't classify is processed again, first, in the next round.

A clean memory every round.

An error of judgment doesn't carry over into the next decision.

Nothing fails silently.

What fails is retried and written down; what's abandoned shows up in the report.

The four products

Everything happens inside Husker.

The four products feed each other inside Husker. Red Team finds the gap, Defense blocks it, Report keeps the evidence and Cisent Iroh turns every case into learning for your team.

Sentinel Red Team

Continuous ethical hacking.

AI agents attack your systems with your authorization and within the scope you define, to find the gaps before an attacker does.

Explore Sentinel Red Team

Sentinel Defense

Agentic defense in real time.

Detects, assesses and blocks attacks alongside your EDR, your firewall and every source you connect. It acts on its own only where an action can be undone, and asks before anything with impact.

Explore Sentinel Defense

Sentinel Report

The report on every attack, while it's still open.

Gathers and seals the evidence and drafts the report for each audience. For any company, whether or not Law 21.663 applies to it.

Explore Sentinel Report

Cisent Iroh

The agent that trains your team.

Talks with your team, explains alerts and incidents and teaches how to respond. It's an add-on to any of the other three products.

Explore Cisent Iroh

Manthys

The models that judge.

Mind reasons, Sentinel defends, Nova brings the scale and Reflex filters. They work inside Cisent alongside Husker, the agentic layer.

Manthys Mind · generalist

It says “I don't know” when it doesn't know.

Explore Mind

Manthys Sentinel · cybersecurity

It reads the attack like an analyst.

Explore Sentinel

Manthys Nova · the largest scale

It flags its doubt. So you can verify.

Explore Nova

Manthys Reflex · decisions

It decides what deserves the judge's attention.

Explore Reflex

Book a meeting with the Cisent team.

Tell us what your organization needs to protect. We'll reply on WhatsApp and set up the meeting.

Message us on WhatsApp and let's book a meeting

or write to info@cisent.ai