You spent the last 12 months looking at MCP servers, scratching your head.
What is the point of these? This is just an API. I can move data from one product to another. I don’t understand it.
You build AI for a living. And you still didn’t get it.
Then it finally clicked. And it has blown your mind ever since.
Neo downloads kung fu
There’s a scene in The Matrix where Neo is plugged into the chair, a training program gets loaded straight into his brain, and he opens his eyes and says, “I know kung fu.”
Morpheus looks at him. “Show me.”
You’d spent a year looking at the cable. You’d missed the program.
That is what you are doing with MCP servers.
MCPs allow you to bundle your IP, your content, your resources, your specific tools, ways of thinking and methods of working into a scalable and distributable system that gives other people’s AI superpowers. Someone plugs it in, and their AI knows your kung fu.
What an MCP server actually is
Here it is in one sentence. An MCP server is a box you stick your best thinking into, which any AI can plug into and use, on rules you set.
MCP stands for model context protocol. Break it down a layer at a time:
- Model. This is just the AI doing the work. Claude, ChatGPT, one of the open source models. It doesn’t matter. Bring your own flavour.
- Context. In the world of AI, context is everything. This is what you have baked into the box: the tools, skill files, resources, IP and knowledge that make the AI good at your kind of work.
- Protocol. The standard set of instructions that makes this universal. So you don’t need to care which AI company is currently on top. This is the adopted standard, and the big AI assistants already support it.
It’s not going to be something that you build that quickly degrades and goes out of fashion.
Your best thinking is stuck in your head
If you run an expert business, you know where your real value lives. In a few heads. Often one.
Every client who wants the good stuff needs that person in the room. Every new client means more of that person’s hours. It doesn’t scale, and the business is only ever as big as the diary.
Your real value lives in a few heads. Often one. It doesn’t scale, and the business is only ever as big as the diary.
So you try to package it. Most people reach for skill files or custom GPTs. If I encapsulate all of my knowledge into these, I can give it to people and they’ll get value from it.
Inside your own business, that’s exactly right. Your team should have your best thinking written down and sitting where their AI can reach it.
The trouble starts the moment you hand it to someone outside.
As soon as you give someone your skill file, there’s no reason to keep paying you for all of that hard work that you have done. They just ride off into the sunset and get the benefit of your work forevermore. Build a custom GPT and the same thing happens. As soon as you transfer it to anyone, you’ve lost control of your IP.
You didn’t sell them a lesson. You sold them the program.
You decide who gets plugged in
With an MCP server, you get to lock your IP in a box, and you can build authentication into that server. Authentication allows you to check… does this incoming AI have my permission to use my IP?
Have they paid me in the last 30 days? Have they paid a one-off fee or a subscription?
If so, the request is accepted and their AI gets the value contained within your server. If not, it doesn’t.
Has this AI paid you? If so, the request is accepted and it gets your best thinking. If not, it doesn’t. You’re renting out the kung fu, not handing over the disk.
Back to the chair. Neo didn’t write the kung fu. Somebody did, and they decided what got loaded and who it got loaded into. That’s you now. Your clients’ AI sits in the chair, your thinking gets loaded in, and you decide who keeps access. Stop paying, and the chair goes quiet.
You’re renting out the kung fu. You’re not handing over the disk.
And they will rent it from you month after month after month.
What goes in your program
This is the plainest way I know to turn your expertise into a digital product. You don’t need to start a big AI company. You don’t need to become a software company.
You need to look hard at what you already own.
If you have an amazing way of building incredible culture in someone’s business, you can package that IP up and give it to the people you work with, so that every time they’re working with their AI, your IP is affecting the output.
Maybe you are in possession of proprietary data sets that are really hard to get and are not available on the general internet. People will pay you a fee to start plugging those capabilities into their AI.
Maybe it’s your methods and mental models. Your frameworks. The chapters of your book. The way you coach a client through their finances, or through a hard conversation with their team.
Bundle it up into a little package that gives other people’s AI superpowers.
And notice what you’ve built. Not a course. Not another hour sold. A product nobody else can make, because nobody else has your head.
They will love you for it. Every interaction they have with their AI is now levelled up by your best thinking. And it is a great way to build connection with your customers over a very long period of time.
It’s wide open right now
The business leaders who see this, while it’s wide open, and move first are going to make out like bandits with AI.
Don’t be sleeping on this.
Write down the one thing your clients only trust you to do. That’s your first program.
Then decide who gets to load it.
Written by