AI Engineering Academy · 강의

MCP란 무엇이며 왜 중요한가

Model Context Protocol이 무엇을 표준화하는지, 임의로 구성한 함수 호출 통합 방식과 어떻게 다른지, 그리고 인공지능 도구 연결을 위한 공통 언어가 되어 가는 이유를 이해합니다.

MCP란 무엇이며 왜 중요한가은(는) CoddyKit의 무료 AI Engineering Academy 강의입니다. 이것은 4개 중 1번째 강의입니다. 아래에서 전체 강의를 무료로 읽을 수 있으며, 내장 코드 에디터와 24/7 AI 튜터와 함께 브라우저에서 직접 실습할 수 있습니다. 이 강의는 AI Engineering Academy 학습 경로의 일부이며, 진행 상황이 웹과 CoddyKit 앱에 동기화됩니다. AI Engineering Academy 강의에는 총 4개의 강의가 포함되어 있습니다.

이 강의의 일부는 아직 번역되지 않았으며 영어로 표시됩니다.

The Tool Integration Problem

Every AI assistant today solves the same problem differently: how do you connect an LLM to your databases, APIs, files, and services? Without a standard, every integration is a custom one-off — OpenAI function calling works differently from Anthropic's tool use, which works differently from Gemini's. Developers rebuild the same adapters over and over.

What Is the Model Context Protocol?

Model Context Protocol (MCP) is an open standard created by Anthropic that defines a common interface for connecting LLMs to external data sources and tools. Think of it like USB for AI integrations — once a tool or data source implements the MCP server spec, any MCP-compatible AI client can use it without custom code. Announced in November 2024, it has rapidly gained adoption across the industry.

The Client-Server Architecture

MCP uses a client-server model: an AI application (the client) connects to one or more MCP servers. Each server exposes capabilities — tools the model can call, resources it can read, and prompt templates it can use. The client (such as Claude Desktop or a custom app) manages the connections and routes tool calls to the correct server.

Client: Claude Desktop, VS Code extension, or your app
Server: A process exposing tools via MCP — a database adapter, a file system, a web search service

MCP's Three Primitives

MCP servers expose three types of capabilities:

Tools: Functions the model can call to take actions, similar to function calling. Example: execute_query(sql).
Resources: Static or dynamic data the model can read — like files, database rows, or API responses. Example: file:///config/settings.json.
Prompts: Reusable prompt templates the user can invoke with parameters. Example: a code review prompt that accepts a diff.

MCP vs. Direct Function Calling

With direct function calling, you define tools in your application code — tightly coupled to your LLM provider's API. With MCP, tools are defined in a separate server process that any client can connect to. Key differences:

Portability: One MCP server works with Claude, custom apps, and future clients
Reusability: Share MCP servers across teams and projects
Separation of concerns: Tool servers can be maintained independently from the AI application

MCP Transport: stdio and HTTP/SSE

MCP servers communicate with clients over two transport mechanisms. stdio is used for local tools — the client spawns the server as a child process and communicates over stdin/stdout. This is simple and secure for developer tools. HTTP + Server-Sent Events (SSE) is used for remote servers accessible over a network, suitable for shared team tools or production deployments.

# Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json)
# Connecting to a local MCP server via stdio:

# {
#   "mcpServers": {
#     "my-database": {
#       "command": "python",
#       "args": ["/path/to/my_mcp_server.py"],
#       "env": {
#         "DATABASE_URL": "postgresql://localhost/mydb"
#       }
#     }
#   }
# }

The MCP Ecosystem

Since MCP's release, a growing ecosystem of pre-built servers has emerged. Anthropic maintains official servers for common integrations, and community servers cover hundreds of services. Rather than building custom tool integrations from scratch, you can drop in an existing MCP server.

Official: GitHub, Google Drive, Slack, PostgreSQL, filesystem, web search
Community: Jira, Notion, Stripe, Snowflake, MongoDB, and many more
Custom: Build your own for proprietary internal tools

MCP in AI Development Workflows

MCP transforms AI coding assistants. With an MCP server connected to your codebase, Claude Desktop can read files, run tests, check git history, query databases, and browse documentation — all from a single chat window. This enables true agentic development where the AI understands and acts on your full development environment.

MCP SDK Overview

Anthropic provides SDKs for building MCP servers in Python and TypeScript. The Python SDK is mcp (install with pip install mcp). You define tools using decorators, resources using URI patterns, and the SDK handles the protocol negotiation, message serialization, and transport details for you.

# Install
# pip install mcp

# Minimal MCP server skeleton
from mcp.server import Server
from mcp.server.stdio import stdio_server
from mcp import types

app = Server('my-first-mcp-server')

@app.list_tools()
async def list_tools():
    return [
        types.Tool(
            name='hello',
            description='Returns a greeting message.',
            inputSchema={
                'type': 'object',
                'properties': {
                    'name': {'type': 'string', 'description': 'Name to greet'}
                },
                'required': ['name']
            }
        )
    ]

@app.call_tool()
async def call_tool(name: str, arguments: dict):
    if name == 'hello':
        return [types.TextContent(type='text', text=f'Hello, {arguments["name"]}!')]

import asyncio
asyncio.run(stdio_server(app))

Why MCP Matters for the Industry

MCP represents a shift toward standardization in AI tooling. Just as REST standardized web APIs and LSP standardized IDE language features, MCP is standardizing how AI models connect to tools. This benefits developers (build once, use everywhere), enterprises (audit and control one integration layer), and the ecosystem (shared, reusable tool libraries).

MCP Adoption and Future Direction

MCP adoption has grown rapidly since its launch. Major coding assistants (Cursor, Windsurf, VS Code Copilot), enterprise AI platforms, and cloud providers have announced MCP support. The protocol continues to evolve with proposals for authentication, sampling, and streaming improvements. Understanding MCP positions you to build interoperable AI integrations that work across the growing ecosystem.

Quick Check

Test your understanding of what MCP is and why it matters.

Lesson Recap

In this lesson you learned: MCP is an open standard that defines a universal interface for connecting LLMs to external tools and data sources, MCP servers expose three primitives: tools, resources, and prompts, and the MCP ecosystem provides pre-built servers for common services so you don't start from scratch. Next up we build our first MCP server in Python and connect it to Claude Desktop.

자주 묻는 질문

“MCP란 무엇이며 왜 중요한가” 강의는 무료인가요?

네 — “MCP란 무엇이며 왜 중요한가” 전체 내용을 이 웹사이트에서 무료로 읽을 수 있습니다. 인터랙티브하게 실습하려면(내장 코드 에디터와 24/7 AI 튜터), CoddyKit PRO로 업그레이드하면 AI Engineering Academy 강의 전체를 잠금 해제할 수 있습니다. AI Engineering Academy 강의에는 총 4개의 강의가 포함되어 있습니다.

“MCP란 무엇이며 왜 중요한가”에서 뭘 배우나요?

Model Context Protocol이 무엇을 표준화하는지, 임의로 구성한 함수 호출 통합 방식과 어떻게 다른지, 그리고 인공지능 도구 연결을 위한 공통 언어가 되어 가는 이유를 이해합니다. 브라우저에서 직접 실행하는 실습 코드로 AI Engineering Academy을(를) 배우며, 24/7 AI 튜터가 강의를 진행하면서 질문에 답변해줍니다.

AI Engineering Academy을(를) 시작하는 데 경험이 필요한가요?

사전 경험은 필요하지 않습니다. CoddyKit의 AI Engineering Academy은(는) 초급자부터 고급 학습자까지를 위해 구성되어 있으므로, 여기서 시작하거나 처음부터 시작할 수 있으며 자신의 속도대로 진행할 수 있습니다. 이것은 4개 중 1번째 강의입니다.

“MCP란 무엇이며 왜 중요한가” 강의는 얼마나 걸리나요?

대부분의 CoddyKit 강의는 약 5~10분이 소요됩니다. 각 강의는 간결하고 인터랙티브하여 꾸준한 진행이 가능하며, 웹과 앱에서 중단한 부분부터 바로 시작할 수 있습니다.

이 AI Engineering Academy 강의에서 코드를 작성하고 실행할 수 있나요?

네. 모든 AI Engineering Academy 강의에는 내장 코드 에디터가 포함되어 있으므로, 브라우저에서 바로 실제 코드를 작성하고 실행한 후 즉시 AI 피드백을 받을 수 있습니다 — 로컬 설정이 필요 없습니다.

이 강의의 모든 강의

← AI Engineering Academy(으)로 돌아가기