Overview

Model Context Protocol (MCP) is an open protocol used to seamlessly integrate LLM (Large Language Model) applications with external data sources and tools. Whether you are building an AI-driven IDE, enhancing chat interfaces, or creating custom AI workflows, MCP provides a standardized way to connect LLM with its required context.

Infrastructure

Model Context Protocol (MCP) adopts a Client-Host-Server architecture, where each host can run multiple Client instances. This architecture enables users to integrate AI functionality into their applications while maintaining clear security boundaries and focus isolation. MCP is built on JSON-RPC and provides a stateful session protocol that focuses on context exchange and sampling coordination between clients and servers.

Communication participants

• Hosts (Hosts) : LLM applications that initiate connections
• Client (Clients) : Connectors within the host application
• Servers : services that provide context and functionality

Detailed explanation of core components

Architecture diagram

Session flowchart

Host (Host)

The host process serves as a container and coordinator.

• Create and manage multiple Client instances
• Control Client connection permissions and lifecycle
• Implement security policies and agree to requirements
• Handle user authorization decisions
• Coordinate AI/LLM integration and sampling