What Is the Model Context Protocol (MCP)? An Interview‑Style Deep Dive

Standard Answer Overview

MCP (Model Context Protocol) is a model context protocol that provides a standardized interface for AI applications to connect to external data sources and tools. For example, an AI app can use MCP to access local files, databases, search engines, calculators, workflows, or specialized prompts, effectively acting as a “Type‑C” interface for AI.

Extended Q&A

1. Difference between MCP and Function Calling

Function calling is an internal capability of LLMs that depends on the model’s returned result, and the return structures differ across models such as ChatGPT and Claude. MCP, by contrast, is an external standard protocol independent of any LLM, allowing different AI applications to connect to the same MCP server.

2. How does MCP ensure security?

MCP itself does not guarantee security at the protocol level. Developers must follow application‑layer best practices such as sandbox isolation, the principle of least privilege, manual authorization, and providing security documentation.

Detailed MCP Analysis

1. Architecture Design

MCP adopts a client‑server architecture. An MCP Host (e.g., OpenClaw, Doubao) manages MCP Clients, which maintain connections to one or more MCP Servers that provide context information. The interaction flow is illustrated in the diagram.

MCP Host: The AI application itself, responsible for managing all MCP Clients.

MCP Client: Maintains the connection to an MCP Server and obtains context for the Host.

MCP Server: Supplies context information to Clients; can run locally or remotely.

2. Three Basic Types

MCP supports three resource types, each with corresponding methods: list (/list), get (/get), and tools/call. The client can, for instance, use the /list method to discover available MCP server types before proceeding with tasks.

Tools: Callable functions such as file operations or database queries.

Resources: Contextual data like file contents or database records.

Prompts: Reusable prompt templates, including system prompts and few‑shot examples.

3. Transport Protocols

MCP Client and Server communicate via JSON‑RPC 2.0 messages and support two transport options:

Standard Input/Output (Stdio): Uses stdin/stdout streams for inter‑process communication on the same machine, offering optimal performance with no network overhead.

Streamable HTTP: Sends messages via HTTP POST and optionally uses Server‑Sent Events (SSE) for streaming. It supports Bearer Token, API Keys, custom headers, and other HTTP authentication methods, suitable for remote server communication.

Note: In March 2025, MCP transitioned from an HTTP + SSE approach to the Streamable HTTP protocol; the reasons for this change are not detailed here.