Agentic AI Master Guide

Introduction

This guide is a master guide on agentic AI, covering as much as possible in this single guide. As such, it will be added to incrementally, and I will write the changelog in the changelog section at the bottom to update on the various changes, so stay tuned for more updates.

Model Context Protocol (MCP)

Picture this analogy. Traditionally, you had two teams working on a particular project. One team spoke French, another team spoke English. You would need to hire a translator that speaks both French and English to carry out communication between them.

Think of an MCP server. The translator? Imagine a puzzle piece which has an indent on both sides. The puzzle piece has an indent on the French side, and an indent on the English side. The connecting puzzle piece between these two pieces is the translator. An MCP functions the same way as this puzzle piece. It connects two different so-called protocols, in this case the agent and various tools that are required.

Why it exists

Traditionally, if a large language model or agent was required to communicate with certain services, for instance, productivity services such as Google’s suite or other tools, API usage would be required. However, not all tools have this API usage. And as such, MCP provides a bridge between the agents and tools.

The three words

MCP splits into three words: model, context, protocol.

• Model: the French-speaking team. They provide the reasoning, intelligence, and task execution capabilities.
• Context: from the tools. For example, your Google workspace or your productivity tool suite. The other part of the puzzle.
• Protocol: the middle connecting one. The bridging puzzle piece that allows communication to go through.

Changelog

• 2026-06-21: First version. Added the Model Context Protocol (MCP) section, incorporated from the standalone MCP explainer.