A while back I was trying to wrap my head around MCP. Not because it's complicat
By Nicholas Hodge · May 27, 2026 · Curated by George's Blog
A while back I was trying to wrap my head around MCP.
Not because it's complicated. Because every explanation out there was just words. And if you're wired like me, words alone don't cut it.
I need to either see it or build it. Visual or hands-on. One or the other, sometimes both, before the pieces actually connect.
So I drew it out. And the second I did, it clicked.
MCP is just a bridge.
Your AI lives in a box. It can think, reason, write. But on its own it can't touch anything outside that box. No databases, no Slack, no live data. Nothing real.
MCP is the door out of that box.
A small connector sits between your AI and whatever tool you need it to use.
The AI says "show me the files in my Drive folder" in plain English. The MCP server translates that into the actual API call Google Drive understands.
Drive responds. AI reads it. Done.
Without MCP: smart chatbot that only knows what you paste into it.
With MCP: AI that reads live data, sends messages, updates records, and takes action across your entire stack.
Most people hear "AI agents" or "agentic AI" and think it sounds complicated.
It's not. You've just given your AI the ability to access tools and take actions automatically. That's it. That's the whole idea.
MCP is one of the main ways you do that.
If you're a visual learner, the diagram below will explain it faster that I did haha