Lessons - MCP Academy

Featured Lessons

Article

Understanding Security in MCP

Security within the Model Context Protocol (MCP) presents a dual challenge. It requires addressing both traditional web vulnerabilities and a new class of threats specific to AI agents. As MCP standardizes how agents interact with external tools and data, it inevitably creates new attack surfaces that demand a comprehensive, defense-in-depth security strategy. Properly securing an …

Article

The Practical Guide to MCP Server Adoption

This framework helps you figure out whether MCP servers are right for your specific use case by systematically checking eight critical constraint categories. Each constraint gets scored, and the combined assessment guides your decision, a true compatibility test for you and MCP. The eight constraint categories What’s included in this framework Constraint assessment matrix For …

Article

The Agentic Intention Framework

Introduction: The paradox of building with AI in the loop As modern software engineers, we’re trained in the agile way: ship fast, iterate based on feedback, let design emerge. We start minimal and evolve based on what we learn from users. But architecting robust, performant, and useful MCP servers demands the opposite: with LLMs in …

Article

The Agentic Intention Framework In Practice

Overview: Who, what and why WHO will use the framework: WHAT are the components of the framework: A structured framework that transforms vague AI agent ideas into precise, measurable workflow definitions by eliciting explicit answers to critical questions before development begins: Framework constraints WHY use the framework: Achieve higher adoption rates for agentic tools and …

Article

Introduction: Terminology Matters

We’ve all seen articles about “an AI system”, but what does that actually mean? The raw LLM that generates text? The agent that makes decisions? The tools it calls? The infrastructure that makes it all work reliably? Without clear terminology, teams build brittle integrations, set wrong expectations, and misplace trust—expecting intelligence where there’s really just …

Article

AI System

An AI system is the comprehensive integration of components that collectively deliver artificial intelligence-powered functionality. It combines the generative capabilities of LLM, the reasoning and decision-making processes of agents, the action-enabling capabilities of tools, and the reliability mechanisms of the orchestration layer. Additional infrastructure components, such as integration layers, plugin frameworks, or standardized protocols like …

Article

AI Agent

An AI agent is a type of AI system that uses a LLM for reasoning and autonomously pursues goals by making decisions, invoking tools, and adapting to feedback. Agents add goal-directed behavior on top of raw model outputs, enabling systems to act in dynamic and flexible ways. Agents can use different mechanisms to access tools, …

Article

AI Tool

An AI tool is a capability beyond an LLM’s native text generation that an AI agent can invoke to access data, perform actions, or integrate with services. Tools can be connected in different ways: through direct API calls, plugin frameworks, or standardized protocols such as the Model Context Protocol (MCP). Key characteristics Also: A tool …

Article

MCP Client

An MCP client is the component of the Model Context Protocol (MCP) that allows AI agents or AI systems to discover and invoke tools or resources in a standardized way. Instead of relying on custom integrations, the MCP client provides a common interface so that agents can call tools exposed by MCP servers dynamically and …

Article

MCP Host

An MCP host is the runtime environment where the AI agent and the MCP client operate. It provides the context in which the agent reasons, invokes tools through clients, and connects to MCP servers. The host is responsible for mediating these interactions, applying permissions, and ensuring that the system runs within a controlled environment. Key …