What is Model Context Protocol?

Model Context Protocol (MCP) is an open standard that defines how AI agents connect to and invoke external tools, data sources, and services. It provides a common interface for tool discovery, tool invocation, and result handling, enabling AI systems to act beyond their training data and interact with live environments in a structured, interoperable way.

MCP was introduced by Anthropic in November 2024 and has since become the de facto standard for connecting large language models to external capabilities. It separates the concern of what a tool does and how to call it from the execution environment in which those calls take place.

How does Model Context Protocol work?

MCP operates on a client-server architecture. An MCP server exposes a set of tools – functions, APIs, or data sources – along with their schemas and descriptions. An MCP client, typically an AI agent or orchestration framework, queries the server for available tools, selects the appropriate one for a given task, and submits a structured invocation request. The server executes the tool and returns results to the client.

Connections use transport protocols such as standard input/output (STDIO) for local integrations and HTTP with Server-Sent Events (SSE) for remote deployments. Within a session, MCP maintains stateful context across multiple tool calls, allowing agents to chain tool use without re-establishing context at each step.

Crucially, MCP defines what tools do and how to call them but makes no requirements about where tool calls run or what state they share with other concurrent calls. This design keeps the protocol lightweight and broadly applicable, while leaving execution context as an implementation concern for each client.

What is Model Context Protocol used for?

MCP is used to connect AI agents to external capabilities such as file systems, code execution environments, databases, APIs, and internal enterprise tools. It is a foundational component in agentic AI pipelines, enabling agents to read data, write to systems, run tests, and interact with cloud infrastructure as part of multi-step automated workflows.

Common use cases include coding agents that execute and test code, knowledge agents that retrieve information from enterprise databases, document agents that read and edit structured content, and infrastructure agents that provision and manage cloud resources. MCP’s open and extensible nature allows it to be adopted across AI frameworks and platforms, making it a shared substrate for the broader agentic AI ecosystem.