MCP (Model Context Protocol)
Last Updated: September 10, 2026 | By Mihail Sebastian | AI Dictionary
An open standard for connecting AI applications to external tools and data through a common client-server interface, introduced by Anthropic in 2024.
What is MCP (Model Context Protocol)?
MCP (Model Context Protocol) is an open standard that connects AI applications to external tools and data sources through a common client-server interface, so any compliant application can work with any compliant server.
Anthropic introduced the protocol in late 2024, and adoption spread across the industry through 2025. Before MCP, every AI application needed a custom integration for every tool it touched; the protocol replaces that many-to-many wiring with one interface.
How MCP Works
An AI application runs an MCP client that connects to one or more MCP servers. Each server exposes capabilities of three kinds: tools the model can invoke, resources it can read (files, database records, documents), and prompts, reusable templates for common tasks.
The client discovers what a server offers, and the model reaches those capabilities through tool calling: it requests an action, the server executes it and returns the result. The same server works with any MCP-compatible application, which is the point of a standard.
MCP and AI Governance
An MCP server is third-party code running with real permissions: file access, API credentials, database connections. Installing one extends what an AI agent can do, and what it can damage, without any change to the model itself.
Organizations respond by cataloging approved servers in an AI registry and routing agent traffic through an AI gateway that logs calls and enforces limits. The governance question shifts from “which model do we allow” to “which tools do we allow it to reach.”
Example of MCP
A support engineer connects an AI assistant to the company’s ticketing system through an MCP server. The assistant’s client queries the server and discovers two tools: search tickets and post a comment.
Asked to summarize open incidents, the model calls the search tool. The server queries the ticketing API with its own stored credentials and returns the results as context for the answer; the assistant never holds the credentials.
Before any of this, the server passed the company’s intake review and entered the registry as approved, with a named owner. A colleague who wants a different server for the same system starts there, not in production.
Related AI terms: Tool Calling · AI Agent · AI Registry · AI Gateway · Large Language Model
Did you like the MCP (Model Context Protocol) gist?
Learn about 250+ need-to-know artificial intelligence terms in the AI Dictionary.
