What is Model Context Protocol (MCP) in AI?

Also called MCP, model context protocol

Definition

Model Context Protocol (MCP) is an open protocol, introduced by Anthropic in November 2024, that standardizes how AI applications connect to external tools, data sources, and prompts through MCP servers.

Model Context Protocol (MCP) is an open standard that defines how AI applications connect to external tools and data. Anthropic introduced MCP on November 25, 2024, to replace one-off integrations between each AI application and each data source with a single protocol. In December 2025, MCP became a founding project of the Linux Foundation's Agentic AI Foundation, which now hosts its development.

How MCP works#

MCP works through three roles defined in the MCP specification: a host (the AI application, such as an IDE or a chat app), clients (connectors inside the host, one per server connection), and MCP servers that expose capabilities. All messages use JSON-RPC 2.0.

An MCP host contains three clients, each connected to its own server: a local files server over stdio and remote tickets and database servers over Streamable HTTP. Host (AI application)Client 1Client 2Client 3Files serverLocal subprocessTickets serverRemoteDatabase serverRemotestdioStreamable HTTPStreamable HTTP
Figure 1: One MCP host with a dedicated client per server connection.

Servers offer three primitives: tools (functions the model can call), resources (read-only data supplied as context), and prompts (reusable templates a user invokes). The standard transports are stdio, for local servers launched as subprocesses, and Streamable HTTP, for remote servers, with an authorization framework for HTTP that recommends OAuth for obtaining tokens. The current revision, dated 2026-07-28, makes the protocol stateless: every request carries its protocol version and client capabilities, replacing the connection-level initialization handshake of earlier revisions.

MCP vs function calling#

MCP and function calling solve different layers of the same problem. Function calling is a model capability: the model receives tool definitions in its request and returns a structured call. MCP is the integration layer: it standardizes how an application discovers tools (tools/list), executes them (tools/call), and connects to the systems behind them.

A host typically reads tool definitions from its MCP servers, passes them to the model as function definitions, and routes each resulting call back to the right server. The practical benefit is reuse: one MCP server for a database works with every MCP-compatible host, instead of each application writing its own integration.

What is Model Context Protocol covers architecture and security in depth, and A2A vs MCP explains how MCP relates to agent-to-agent protocols. The MCP topic hub collects related guides.

Cite this definition

The AI Engineering Editors. "What is Model Context Protocol (MCP) in AI?" The AI Engineering, updated Nov 06, 2026. https://theaiengineering.com/glossary/model-context-protocol

Title
What is Model Context Protocol (MCP) in AI?
Author
The AI Engineering Editors
Publisher
The AI Engineering
Updated
View as Markdown

Part of the Model Context Protocol hub · Updated