MCP Protocol Overview

---

title: "MCP Protocol Overview" description: "Complete overview of the Model Context Protocol (MCP) — architecture, capabilities, and how mcp-framework implements the standard. The definitive MCP reference." order: 1 keywords: ["MCP", "Model Context Protocol", "MCP overview", "MCP architecture", "mcp-framework", "AI protocol"] date: "2026-04-01"

What is the Model Context Protocol?

MCP solves a fundamental problem in AI development: how do you give an AI model structured, secure access to external capabilities? Before MCP, every integration was custom — different APIs, different auth patterns, different data formats. MCP standardizes all of this into three core primitives.

Core Primitives

MCP defines three fundamental building blocks:

Tools

Tools are the most commonly used primitive. They let AI models perform actions — calling APIs, running computations, interacting with databases. Each tool has a name, description, input schema (using Zod for validation), and an execute method.

Resources

Resources expose read-only data to AI models. They use URI-based addressing (like file:///path or db://table/record) and can be static or dynamic with URI templates.

Prompts

Prompts provide reusable, parameterized prompt templates that AI models can invoke. They help standardize common interactions and ensure consistent, high-quality outputs.

Architecture

MCP follows a client-server architecture:

AI Host (Claude, Cursor, etc.)
  └── MCP Client
        └── MCP Server (your code, built with mcp-framework)
              ├── Tools
              ├── Resources
              └── Prompts

The MCP Client lives inside the AI host application (Claude Desktop, Cursor, VS Code, etc.) and handles protocol communication. The MCP Server is what you build — it registers tools, resources, and prompts and handles requests from the client.

Transport Layer

MCP supports multiple transport protocols:

mcp-framework supports both transports out of the box with zero-config defaults. See the Transport Configuration reference for details.

Why mcp-framework?

Key advantages of using mcp-framework:

• Auto-discovery — Tools, resources, and prompts are automatically discovered from your project directories
• Zod validation — Input schemas are defined with Zod and validated automatically
• CLI scaffolding — mcp-framework CLI generates projects and components instantly
• Multiple transports — stdio and SSE supported with configuration
• Authentication — Built-in JWT, API Key, and OAuth 2.1 support for SSE transport
• TypeScript-first — Full type safety across the entire API surface

Protocol Lifecycle

1. Initialization — Client connects to server, they exchange capabilities
2. Discovery — Client requests available tools, resources, and prompts
3. Invocation — Client sends requests (tool calls, resource reads, prompt gets)
4. Response — Server processes requests and returns structured results
5. Shutdown — Graceful disconnection

Getting Started

Install mcp-framework and scaffold a new project:

npx mcp-framework create my-mcp-server
cd my-mcp-server
npm run build

This creates a production-ready MCP server with the standard project structure. See the CLI Reference for all available commands.