Steven Gonsalvez

Software Engineer

← Back to Blog

Model Context Protocol (MCP) Series

4 parts in this series

1

Introduction to Model Context Protocol (MCP): The USB-C of AI Integrations

Model Context Protocol (MCP) explained for developers 2025. How Anthropic MCP standardises AI tool integrations, replaces LangChain connector chaos with M+N simplicity. MCP servers, clients, and the USB-C analogy.

2

MCP Architecture Explained: STDIO, SSE Transport and What Makes It Tick

MCP architecture deep dive 2025. How Model Context Protocol works under the hood: function calling, JSON-RPC, STDIO vs SSE transport, streamable HTTP, OAuth 2.1 auth, and the enterprise security gaps nobody talks about.

3

Build Your Own MCP Prompt Server: A Dev-Centric Registry with STDIO

Build your own MCP prompt server tutorial 2025. Step-by-step TypeScript guide to a layered prompt registry with STDIO transport, Zod validation, file-based storage. Works with Claude Desktop, Cursor, and any MCP client.

4

MCP Security Risks: Tool Poisoning, Shadowing Attacks and How AI Gets Exploited

MCP security vulnerabilities explained 2025. Hands-on demos of tool poisoning, cross-server shadowing attacks, token theft, and data exfiltration via Model Context Protocol. Practical defences and security best practices for AI agents.