MMOOZZGG.com

Glossary

What is the Model Context Protocol (MCP)?

In short

The Model Context Protocol (MCP) is an open standard that lets AI applications connect to tools and data: files, a CRM, databases and online services.

AI agents · Model Context Protocol

MMOOZZGG.com editorsChecked: 1 min read

For a child

Different devices used to have different chargers, which was a nuisance. Then people agreed on one plug for all of them. MCP is such a common plug for neural networks: through it, AI can connect to different programs in the same way.

In plain words

For AI to work with your programs, people used to build a separate connection for each pair of neural network and program. MCP agrees on one way for all of them. A program makes an MCP server once, and any AI app that understands MCP can use it: read data and run actions.

For business

MCP lowers the cost of integrations: a connection to a CRM, 1C (a Russian accounting and ERP system) or a knowledge base is built once and works with different models. This reduces dependence on one AI vendor. The risks are clear too: an MCP server gives the model access to data and actions, so you need to limit permissions and check third-party servers before you connect them.

For a specialist

Anthropic described the Model Context Protocol in November 2024 as an open specification. It uses a client-server design: the host (an AI application) holds MCP clients, and each connects to an MCP server over JSON-RPC 2.0 through stdio or Streamable HTTP. The server exposes primitives: tools (callable functions with input schemas), resources (data for context) and prompts (templates). At the server’s request, the client can ask the user for data or confirmation (elicitation); sampling is marked deprecated in the 2026-07-28 version of the specification. Since December 2025 the project has been part of the Agentic AI Foundation at the Linux Foundation. Security rests on authorization, explicit user consent to calls and isolation of servers; the main threats are prompt injection through tool data and excessive permissions.

Example

A company connects an MCP server for its CRM. Now, when a sales manager asks the AI assistant to “show deals with no reply for more than three days”, it gets the list from the CRM and drafts the emails.

Related terms

Questions and answers

How is MCP different from an API? Each program has its own API. MCP is a common standard on top of them, so AI applications find and call the tools of different programs in the same way.
Who supports MCP? The standard is open, and many AI applications and developer tools support it. The official protocol site lists examples of apps with MCP support.

Sources

Where companies use it

Not sure whether your company needs this?

AI solutions map
Contact us