← Terminal to Agents
Terminal to Agents

What MCP Actually Is — a Shared Plug Shape

21 Sept 2026·7 min read

Last part ended on a gap: an agent in your terminal can only reach things that have a command on your machine. Your ticket tracker, a hosted database, your company wiki — those live behind a server and an API, the two words from Part 9. Getting an AI to them was, until recently, a genuinely bad problem.

The problem was multiplication

Say there are five AI apps and twenty systems worth connecting. That is not twenty-five pieces of work. Each app needs its own custom glue for each system, so it is five times twenty — a hundred. And every one of those hundred separately reinvents signing in, handling failures, and describing what the tool can do. Badly, usually, because nobody has time to do a hundred of anything well.

This is why so many obvious integrations never existed. Not because they were hard, but because a hundred pieces of near-identical work is nobody’s job. MCP replaced the multiplication with addition: agree on one shape, and each side builds to it once.

Where the name comes fromModel Context Protocol. A protocol is only an agreement about shapes — like the shape of a plug. The plug does not know what a kettle is, and the kettle does not know what a power station is. That is the entire point of a standard: neither side has to know the other.

Two sides, plainly

  • The client is the AI app you are sitting in front of. Your coding agent, a desktop assistant. It knows how to speak MCP and nothing about your particular systems.
  • The server is a small program that exposes exactly one system — your files, a database, a ticket tracker. It knows that system properly and nothing about which AI is asking.

Server here has nothing to do with a data centre. It is usually a small program running on your own laptop. Same word as Part 9, same plain meaning: something that waits to be asked and answers.

What a server offers: three things

  • Tools are named actions the model can take — create a ticket, run a query. Each one carries a description the model reads and a declared list of what it needs, so the model can discover what is possible instead of being told in advance.
  • Resources are things to read, sitting at an address — a file, a record, a page. They inform an answer; they do not change anything.
  • Prompts are good instructions saved next to the tool, so the sensible way to use it travels with it instead of living in one person’s head.
The rule of thumb that settles itIf it changes something, it is a tool. If it only tells you something, it is a resource. That one line resolves most of the confusion people have when they first read the spec.

Notice what the tool description is doing. It is the same idea as --help in Part 5, pointed at a model instead of a person: the tool explains itself, so nobody has to hard-code a list of what exists. That is why a client written last year can use a server written tomorrow.

Adding one, in practice

claude mcp add --transport http sentry https://mcp.sentry.dev/mcp
A remote server over HTTP. This exact example is printed by claude mcp add --help on this machine.
claude mcp add my-server -e API_KEY=xxx -- npx my-mcp-server
A local one that runs as a program. Everything after the two dashes is the command to start it.
claude mcp list
Shows what is configured and whether each one is actually connected.

That last command is worth running before you debug anything, because it tells you the difference between a server you configured and a server that is answering. Mine lists several, and some of them say they need authentication — which is the subject of the final part.

One thing most tutorials still teach wrongThe current spec removed the session handshake. There is no initialize step and no session id any more — every request carries its own version information instead — and the older HTTP+SSE transport is deprecated. Plenty of material still walks you through that handshake as if it were required, which is why some tutorials no longer match what the tools actually do. If a guide opens by setting up a session, check its date before following it.
Do this in 60 secondsRun claude mcp list. If it prints nothing, that is a correct answer — you have none configured yet. Then run claude mcp add --help and read the transport options. You are reading a tool describe itself, which is the same habit as Part 5.
Today’s AI angleMCP is not an upgrade to the terminal, and this is where people go wrong. An agent with a shell already has every installed tool. MCP answers a different question: what about the systems that were never on your machine to begin with? Knowing which question you have is the whole of the final part.

Which brings us to the finale, and to the most over-built thing in AI right now: an MCP server for something a plain command already does perfectly well.

What MCP Actually Is — a Shared Plug Shape · Ali Huzaifa