(I’m misleading you)
Yesterday, I “asked” ChatGPT to create my five TikToks and YT-Shorts for the day.
Five minutes later, five videos: scripted, titled, tagged, visuals and audio done, scheduled for the best posting times.
I didn’t explain anything. It just… knew what I needed.
Scan the social chatter > pick a topic I haven’t covered > craft a take in my voice > make the script > build the visuals > produce the audio > generate the descriptions, tags, etc. > scheduled.
Boom. Done.
Pretty wild, right? Except… it’s not real.
I made that up.
(Told you I’m misleading you)
ChatGPT can’t do all of that natively. Not yet. I think it will. But not yet.
But here’s the kicker: you can make this happen today. No coding. An afternoon’s work.
And the reason is something called Model Context Protocol (MCP).
DON’T ZONE OUT ON ME. KEEP IT TOGETHER.
I get it. MCP sounds like engineer-stuff. Google Scholar stuff. “Do I even have the credentials to read about this?” (Yes.) “Do I need my college ID still?” (No.)
If you’ve ever thought, “I wish I could actually use these AI tools better than just talking about intermittent fasting with ChatGPT… but I can’t keep up”…
Then you should know about MCP.
MCP is how AI stops just talking to you, and starts actually doing things for you.
It’s the difference between asking your friend for advice… and giving them the keys to your house so they can just go fix the problem for you. (Dropped your crown, queen 👑)
Here’s the thing, you need agency if you want agents.
Most people approach AI, subconsciously with a mindset like this:
“How can I adopt the cool things other people built?”
The real unlock is to think like this:
“How can I make cool things for myself?”
You don’t have to be a developer. But you do need to know which Lego pieces exist and become willing to learn how to snap them together.
MCP is one of those pieces.
Except the lego metaphor is actually kind of bad…. Think of it more like the universal remote that lets your AI plug into anything: data, tools, automations, services.
Here’s the plain-English version:
MCP is a standard for letting AI connect to other stuff.
That “stuff” could be:
Whatever database you have
Your SaaS tools
An automation platform like Zapier or n8n
A live feed of market data, weather, social posts, etc.
Pretty much anything with an API
The “MCP server” is the shop that offers those connections.
The “MCP client” is your AI walking into the shop and saying, “I’ll take two of those.”
Once connected, your AI can actually do things, not just tell you what it would do.
Want to get more technically oriented? Anthropic has a great intro to MCP (they also own Claude), they even built a site for it, and PulseMCP is a great resource for exploring what MCP capabilities are available to you.
Why You Should Care
Because this is where the “AI superpowers” everyone talks about actually start to show up. It’s the “a-ha” moment you might be waiting for.
Get past the smell of ‘too technical’, dig into it a little, try it out and get the unlock.
And you can set this up yourself today. This isn’t something you need a course on.
Last week, I filmed a real-time, start-to-finish setup connecting Notion to Claude desktop via MCP and it took 10 minutes. No coding. Just clicks. Watch it here in my newsletter.
If you start playing with MCP now, you’ll skip the consumer-training-wheels phase and jump straight to building AI that does things for you.
Once you do, you’ll stop asking “What can I use?” and start asking “What can I make?”
That’s where the fun (and the advantage) is.
CG
