Hi friends! If you've spent any time around AI developers lately, you've heard three letters over and over: MCP, the Model Context Protocol. People call it "the USB-C of AI." Today we'll unpack what that means — no jargon required.

The mess before MCP

AI assistants are much more useful when they can reach your real tools and data — your database, your files, your calendar, your issue tracker. But before MCP, every connection was a one-off. Want your AI app to talk to a database? Write custom glue code. A different AI app, same database? Write it again. Every AI app multiplied by every tool meant a pile of duplicate integrations — developers call this the M × N problem: M apps times N tools equals M × N custom connectors. Painful, fragile, and nobody's favorite job.

The USB-C analogy

Remember when every gadget had its own charger, and your drawer was a graveyard of incompatible cables? Then USB-C arrived: one standard plug, and suddenly any charger works with any device. Neither side needs to know anything about the other beyond the standard.

MCP does the same for AI. It's an open standard that defines one way for AI applications to plug into tools and data sources. Build a connector once, and it works with every AI app that speaks MCP. M × N shrinks to M + N — each side just implements the standard once.

How it works, in plain words

There are only two roles to learn:

  • MCP servers sit in front of a tool or data source and expose what it can do in a standard format. A database server might expose a run_query tool; a file server might expose read_file and search_files. Servers can also share resources — documents and data the AI can read for context.
  • MCP clients are the AI apps — chat assistants, coding tools, agents. A client can connect to any MCP server, ask "what can you do?", and get back a menu of tools it can call.

That discovery step is the quiet superpower. The AI app doesn't ship with built-in knowledge of your database — it connects, asks, and learns the menu on the spot. Plug in a new server and your assistant gains new abilities, no code changes needed.

A concrete example

Say you ask your AI assistant: "Find our five biggest customers and save a summary to customers.md." With two MCP servers connected, the flow looks like this:

  1. The assistant sees a database MCP server offering run_query, and a file MCP server offering write_file.
  2. It calls run_query to fetch the top five customers by revenue.
  3. It writes a short summary of the results.
  4. It calls write_file to save customers.md to your project.

One request, two completely different systems, zero custom integration code. That's the whole pitch.

Why the ecosystem exploded

Because anyone can build a server, everyone did. There are thousands of public MCP servers now — maintained by tool vendors, companies, and hobbyists — covering just about anything you'd want an assistant to touch:

DatabasesFile systemsGit & GitHubWeb searchBrowsersSlack & emailCalendarsIssue trackersCloud servicesPayment tools

This is the network effect that made USB-C win, playing out again: every new server makes every MCP client more capable, which attracts more clients, which attracts more servers.

Tip: Only connect MCP servers you trust. A server can read what you share with it and act on your behalf, so treat adding one like installing an app — check who published it first.

What should a beginner take away?

Three things. First, MCP is plumbing, not magic — it standardizes how AI apps and tools talk, nothing more. Second, you don't need to build anything to benefit: modern AI apps let you plug in existing servers with a few lines of configuration. Third, when you are ready to build, writing a small MCP server is one of the friendliest AI projects there is — official libraries handle the protocol, and you just describe your tools and write ordinary functions.

One standard plug, endless devices. Once you see MCP that way, half the AI news suddenly makes sense. Next, see how assistants actually use these tools in a loop — read what are AI agents? — and keep going!