Glossary
What is an MCP server?
In short
An MCP server is a program that gives an AI application access to one system through the Model Context Protocol (MCP): it tells the AI which actions and data are available and carries out the model’s requests.
AI agents · MCP server
MMOOZZGG.com editorsChecked: 2 min read
For a child
Imagine a library where each shelf has its own librarian. One knows everything about fairy tales, another about dinosaurs. You do not dig through the books yourself; you go to the right librarian, and they bring you what you are allowed to take. An MCP server is like such a librarian for a neural network: it looks after one shelf and hands out only what is allowed.
In plain words
Suppose a company has a CRM, email and a folder of contracts. For an AI assistant to work with each of them, you place a small go-between program next to each system. It tells the assistant what it can do: find a client, show a deal, create a task. When the assistant needs an action, it calls the go-between, which carries it out in its system and returns the answer. The go-between can run on your computer or on the vendor’s side on the internet. Usually one such server covers one system.
For business
Many software vendors already release ready servers, so some connections need no development. For your own systems, such as 1C (a Russian accounting and ERP system) or an industry database, a server is written once, and different AI applications use it. The cost depends on whether a ready server exists, how many actions you need to open and which permission checks to build in. The main security question: the server works with the rights you gave it, and the descriptions of its tools go straight to the model. A third-party server from an open catalog may ask for extra access or hide someone else’s instructions in its descriptions. Before connecting, check the author, the list of actions and the call log, and leave writes to the books to an employee. Such connections to 1C and the CRM are worked out at the integration stage.
For a specialist
An MCP server is a program that provides context and actions to MCP clients. The host, such as an AI application or a development environment, creates a separate client for each server, and each pair has its own dedicated connection. The server exposes three kinds of primitives: tools, which the model calls; resources, data by URI that the application selects; and prompts, templates that the user starts. Messages use JSON-RPC 2.0: the client gets the list of tools with the tools/list method and makes a call with tools/call. A local server usually runs over the stdio transport and serves one client; a remote one works over Streamable HTTP and serves many, and the documentation recommends OAuth for access tokens. Tool inputs are described with JSON Schema, and the application may ask for user consent before a run.
Example
An accounting team connects an AI assistant to a server for its accounting system with only two actions: find an invoice and show payments by counterparty. The server cannot create documents, so the assistant will say whether a client has paid but will change nothing in the books.
Related terms
Questions and answers
Sources
Where companies use it
Not sure whether your company needs this?
AI solutions map