Sigilant · Control for AI agents
AI agents are being switched on across your business faster than anyone can count them, and they already raise orders, change records and move money. Sigilant gives every agent its own identity, an owner and your rules, so you can say yes to AI and prove it is under control.
Does this sound familiar?
Today, most businesses cannot answer them. With Sigilant, you can, and you can prove it.
The audit
Today: a spreadsheet, if you are lucky. Some agents use a staff member's login. Nobody is sure what each one can reach.
With Sigilant: every agent has its own ID and a named owner, who checks in regularly to confirm it is still needed. When an agent works for someone, the record names both.
When something goes wrong
Today: all the agents share one password, so your systems cannot tell them apart. Nobody can say which one did it, or whether anyone agreed.
With Sigilant: every action is on the record: which agent, for whom, and who said yes. When an agent placed a $48,000 order, it waited for a named manager's approval, and the record shows it.
Your customer
Today: a paragraph of good intentions. Often the AI holds a password to the system, and nobody can say exactly what it can reach.
With Sigilant: the AI never holds a password. Sigilant lets it in for one allowed job, then locks the door again, and you can show your customer the rules and the record.
What it costs you today
Each AI agent needs its own ID, like a staff badge: which agent it is, who it works for, and what it may do. Without one:
Who did it, for whom, and who said yes? Nobody can tell you.
Switching off a shared password breaks everything that uses it. So nobody does, until it is too late.
AI comes built into the software you buy and the tools your teams make. The ones you cannot see are the risk.
Without a safe way to say yes, the answer is no, and the business waits.
See it working
From our demo: an AI agent that buys things, a manager called Priya, and a finance system. The rules for what the agent may do, and who must approve it, live in Sigilant, not inside the agent or the finance system.
How it works
Your AI agents go through Sigilant to reach your systems, instead of going straight in. Nothing has to be rebuilt.
Step 01
Before anything happens, Sigilant checks it against your rules: which agent, for whom, doing what.
Yes, wait or no
Step 02
Big or risky actions wait for a named person to say yes. They see exactly what they are agreeing to.
A person decides
Step 03
The AI never holds the keys to your systems. It is let in for one job, then the door locks.
Nothing to steal
Step 04
Everything is written down, yes or no, on a record that shows if anyone changes it.
Proof, automatically
What you get
Each one is something Sigilant does, in words you can repeat.
Each one has its own ID and a named owner.
Every action is checked as it happens.
Big or risky actions wait for approval.
No more than its job, or its person, allows.
Let in for one job, then the door locks.
Stop any agent at once, even halfway through.
Nothing it reads can change your rules.
A record nobody can quietly change.
Modules
Start with control over your AI agents. Add finding and testing, and the same control for plant and machinery.
Available
Control for AI agents
Every agent has an ID and an owner, follows your rules, waits for a person's yes on big actions, and can be switched off at once.
Available
Find your AI, and keep testing
Finds the AI you did not know you had, then keeps testing your systems for weaknesses, every day and after every change. Its own testing agents work under Sigilant's rules too.
In development · design partners wanted
Control for plant and machinery
No AI agent can change a controller without your rules and a named engineer's yes. Safety systems are out of reach, always.
In development
Test the plant without risking it
Only listens on live sites, does active testing in a lab, and makes a live check only when two people approve.
Why OT, why now
Built for Australian water, energy and critical infrastructure under the SOCI Act. Runs fully offline, on site.
Become a design partner
cisoai.au · from The Register · 11 Oct 2026
AI Agents Reached Industrial Controls in Minutes During Booz Allen Lab Tests
In Booz Allen Hamilton's lab, AI agents went from the edge of the network to acting inside an industrial control network in just over 16 minutes, and moved a robotic arm.
Read the brief →
Deployment
Your people log in the way they already do, with Okta, Auth0 or Microsoft Entra ID. Sigilant runs on your own systems, so your rules and your records stay with you.
In one hour, we will show Sigilant controlling it, on your systems, with your rules. You leave with the record.
Get in touch
Tell us which AI agents you want to use and what they need to reach. We will reply within one business day.
