Model Context Protocol (MCP)

Overview

The Model Context Protocol (MCP) is the foundational layer for connecting AI agents to the external world. Drawing inspiration from the Language Server Protocol (LSP), MCP defines how AI applications communicate with servers that expose external capabilities.

Why MCP Matters

Before MCP, every AI integration required custom code. Each combination of AI platform + external service meant a new integration. MCP changes this by providing a universal interface—build one MCP server, and it works with Claude, ChatGPT, Gemini, Copilot, and more.

Architecture

MCP uses a client-server architecture:

• Hosts: Applications like Claude Desktop, Cursor, VS Code that run MCP clients
• Clients: Components that maintain connections to MCP servers
• Servers: Programs that expose capabilities through the standardized protocol

Core Capabilities

Tools

Executable functions that agents can invoke, described with JSON Schema for automatic parameter validation.

Resources

Access to data sources—files, databases, APIs—with support for real-time subscriptions.

Prompts

Reusable prompt templates for consistent interactions.

Transport Options

• stdio: Local process communication (most common)
• HTTP/SSE: Web-based deployments
• WebSocket: Real-time bidirectional

Linux Foundation Governance

In December 2025, Anthropic donated MCP to the Linux Foundation as part of the Agentic AI Foundation (AAIF). Founding members include OpenAI, Google, Microsoft, Amazon, AWS, and Bloomberg.

This ensures MCP remains:

• Vendor-neutral
• Community-driven
• Open source

Adoption Scale

• 97M+ monthly SDK downloads
• 10,000+ active MCP servers
• First-class support in ChatGPT, Claude, Gemini, Copilot, VS Code, Cursor

Security Considerations

Known vulnerability patterns:

• Prompt injection via tool responses
• Tool permission escalation
• Lookalike tool attacks

Best practices: sandbox execution, validate inputs, principle of least privilege.

When to Use MCP

Use MCP for:

• Connecting agents to APIs, databases, file systems
• Building tool ecosystems
• Standardizing integrations

Don't use MCP for:

• Agent-to-agent communication (use A2A)
• User interface streaming (use AG-UI)

// TypeScript MCP Server Example
import { Server } from "@modelcontextprotocol/sdk/server/index.js";

const server = new Server({
  name: "weather-server",
  version: "1.0.0"
}, {
  capabilities: { tools: {} }
});

server.setRequestHandler("tools/list", async () => ({
  tools: [{
    name: "get_weather",
    description: "Get current weather for a location",
    inputSchema: {
      type: "object",
      properties: { location: { type: "string" } },
      required: ["location"]
    }
  }]
}));