Relatively newer AI agents based on large language models (LLMs), such as GPT-4o, Claude, or Gemini, are very proficient at general reasoning and answering broad questions. However, they usually struggle with domain-specific queries—like “Give me only Dell devices information”—because they don’t have access to proprietary, internal, or organization-level data.
To answer these kinds of questions correctly, an LLM requires more than just a prompt: it needs contextual information made available through trusted internal sources. This article shows you how to build an AI agent that can access and use domain-specific context, thanks to the Model Context Protocol (MCP). It includes a code example of custom MCP Server creation and demonstrates how an MCP Host (in this case, VS Code) talks to the server and uses a Tool.