TL;DR: Model Context Protocol (MCP) proposes a shared standard like USB for AI that lets assistants access relevant context directly from your apps.
Remember that one friend who turns finding a lost item into a twenty‑question interrogation?
AI assistants have become incredibly helpful -- but sometimes, they feel just like that friend, especially as our expectations keep growing.
You: When was my 1:1 with Priya moved to?
LLM: I don’t have access to your calendar or email. Could you: – Upload your calendar file? – Paste the reschedule email? – Tell me when it was originally scheduled?
You: I think it was last Thursday. Priya mentioned the change in Slack.
LLM: Can you paste that message here?
You:
What if your AI assistant could channel its inner Sherlock...
You: When was my 1:1 with Priya moved to?
LLM: Checking your connected tools... ✅ Found reschedule email from Priya in Gmail — moved from Thursday to Friday ✅ Confirmed new event in Google Calendar: Friday, 11:30 a.m. ✅ Cross-referenced Slack chat from Tuesday confirming the change
Your 1:1 with Priya was rescheduled to Friday at 11:30 a.m., confirmed by calendar, email, and Slack, Elementary!. Would you like me to send you a reminder?
Imagine a world without USB. Every PC manufacturer would have its own ports for connecting devices like keyboards and mice.
Peripheral makers would need to build multiple versions of the same device to support different ports. That’s exactly what happens with electrical plugs today as each country has its own socket type.
What is Model Context Protocol (MCP)?
Model Context Protocol(MCP), like any other protocol, is a set of rules, kind of like the USB standard for computers, that aims to help AI assistants connect to apps(email, todo / task apps etc) and data. Think of a protocol as a shared language that allows different things to understand each other.
To be fair in comparison, USB devices do need computers to work and vice-versa — it's a symbiotic relationship.
That's not the case with the apps as they can function independently.
Why does it matter?
Just like Apple or Android can’t build every app and Slack can’t build every bots, AI assistants can’t build every capability users need. They must rely on the wider developer community to extend their capabilities by adapting existing apps to work with AI assistants through the MCP architecture.
When should you use it?
It's still early days for MCP, like when USB was first introduced. Not all computers or devices used it right away. Similarly, for MCP to become really useful, both the companies that make AI assistants (like the "computer" in the USB analogy) and the companies that make apps (like the "keyboard" or "mouse") need to adopt it.
For end-users:
You can add more capabilities to your favorite AI assistant. These are referred to as MCP servers—a somewhat confusing name, as they function more like extensions to your AI assistant. MCP is still in limted beta and is aimed mainly at developers.
For developers:
Should you jump on the bandwagon of building an MCP Server—in other words, making your app’s capabilities available to be used in the AI assistant client?
That depends. Ask yourself: how long your product can stand out if users only access it through an AI assistant.
As a developer, you build an extension to AI assistants aka MCP Server. The MCP Host / Client is your AI assistant, like Claude or ChatGPT. Here is a full list of supported clients.
💬 Your Turn — What Do You Think?
Do you see MCP as a useful standard or just another layer of complexity?
What’s one app you wish your AI assistant could “talk to” seamlessly?
Would you ever build (or use) an MCP Server?
Would you trust an assistant that automatically pulls info from your apps?
Comment below or DM me—I’d love to hear your take.
📚 Next Topic?
I’m on a mission to make tech as easy as ABC! Is there a topic — anything from ABC to AI — that you'd love simplified? 👉 Drop it in the comments, or send me a quick DM, and I’ll break it down in a future post!
✏️ P.S. I’m always tweaking these to make them better — was this one clear, helpful, or missing something? Let me know in the comments or send me a quick DM!
