About
We check that AI does only what it should.
Virasai AI works with startups and small businesses putting AI agents to work. We review, build, secure and monitor agentic AI systems, and we check them with fixed rules rather than asking anyone to take our word for it.
What we stand behind
Our work is in public, so you can check it.
A business like ours asks you to trust its judgment. So we publish what that judgment is built on, including the parts that are unfinished.
- Published research. Our methods are informed by VaHive Systems Lab, an independent research lab. Its preprints are on The Lab, each with a permanent DOI and its own stated limits.
- Open-source tools. Sentinel and Magus OpenSecMCP are free and open source. Open Source describes both, including what each one does not do.
- A real deliverable, published in full. The Adversarial Issues Register is three independent reviews of our own gateway, with every finding, including the ones still open.
- Our method, running live. Watch checks public open-source packages against fixed rules every six hours and publishes what it finds.
How we work
Three commitments that don’t change with the client.
- Checks run on fixed rules. A rule either matches or it doesn’t, and nobody can talk it round, including us.
- An honest verdict. We will say “not yet” or “not worth it” when that is the truth, and tell you when you don’t need us for anything more.
- You keep what we produce. Plans, findings and reports are yours, whether or not you work with us again.
The business
Where we are, and how to reach us.
- Established in South Australia, Australia. We work with clients anywhere.
- ABN 69 780 392 821
- Contact [email protected], or the contact form. We read every enquiry ourselves and usually reply within a couple of business days.
Six multiple-choice questions, about two minutes. You get an answer straight away: a good fit, worth a real conversation, or not a fit right now. We only call if you ask us to.