Why I co-founded KAiZAI

I’ve spent 15 years in corporate IT. If there’s one thing that job teaches you, it’s how hard it is to govern what users do and to protect internal company data while still letting people get their work done.

In March of 2026 I was watching AI agents go from something people talked about to something people were actually connecting to their work. And I kept coming back to the same thought: something has to give corporate IT, security, and compliance teams a way to monitor and control what those agents are being connected to.

That’s what KAiZAI is. It’s the tool I know I’d need in place yesterday if I were the one responsible for AI usage and governance at a company. Which, in my line of work, I usually am.

The problem as I saw it

An AI agent on its own is one thing. An AI agent connected to your email, your ticketing system, your source code, and your documents is something else entirely. Those third-party integrations are where the value is, and they’re also where the risk is.

Anyone who’s worked in IT has seen this before with other tools. People find something useful, they start using it, and IT finds out afterward. The difference this time is how much access is involved and how quickly it’s happening.

The usual answers weren’t going to work. You can block it, and people will find a way around you. You can pick one AI tool, lock it down, and tell everyone that’s the one they get. But the tools are changing too fast for that, and different people do their best work with different ones.

I wanted a third option: let people use whatever AI tool they want, and put the control somewhere else.

One place to connect

KAiZAI is an MCP security gateway. It sits between the agents and the services they want to reach.

It’s agent-agnostic. You can connect to it from any agent. That means there’s no need to configure individual connectors for every AI tool in the company. You set it up yourself, or you tell your users to connect to your single KAiZAI instance, and that’s it. One place, no matter which tool they prefer.

That solves two things at once. For the people using AI, it’s easier: one connection gets them everything they’ve been granted. For IT, it scales: there’s one thing to manage instead of a different set of integrations per tool, per team, per person.

And because everything runs through one place, you can see it and you can control it.

Control that goes down to the tool

Connecting an agent to a service shouldn’t be all or nothing.

Every third-party MCP connection in KAiZAI can be controlled down to the specific tools it exposes. Those controls can then be applied to groups or to individual users. So one team can have a connection with everything turned on, another team can have the same connection with only a handful of tools available, and one specific person can be the exception either way.

That’s the level of control I’d expect as the person accountable for it. “Can our people use AI with this system?” is rarely a yes or no question. It’s usually “yes, for these people, for these things.”

Building it

I’m not a developer by trade. I used AI as a tool to build the SaaS platform itself.

What I do know is corporate IT and security. I know what I’d want to see from a platform like this from both of those viewpoints, and I built around the features I’d expect if I were the one evaluating it. Every decision started from the same question: if someone handed me this product at work, what would I look for first?

The hardest part

The integration marketplace has been the hardest piece.

It’s getting easier with time. More and more vendors are shipping their own native MCP connectors, which simplifies what we need to do on our end. But not every service has one, and custom integrations are still necessary. Those are what take the most time to set up and configure.

Where it is now

KAiZAI is live. If you’re the person at your company who’s going to be asked how AI usage is being governed, or you already have been, take a look at kaizai.io.