Weekly brief
GLOSSARY

MCP (Model Context Protocol)

An open standard that lets AI assistants connect to outside tools and data, such as files, databases or a CRM.

Updated

How does MCP work?

MCP stands for Model Context Protocol. It works through two parts. An MCP server exposes a data source or tool, and an MCP client, built into an AI application, connects to it. The official MCP site compares the standard to a USB-C port: one common way to plug AI applications into outside systems, instead of a custom connection for each one.

Anthropic introduced MCP on 25 November 2024 and released ready-made servers for services including Google Drive, Slack, GitHub and Postgres. Other AI assistants and developer tools have since added support.

Why does MCP matter for a small business?

Each MCP server you connect gives an AI assistant a new reach into your files, systems or accounts. That is useful, and it is also a risk. A local server runs on your machine with the same privileges as the AI application, so installing one is like installing any other program.

Treat MCP servers as software that needs vetting: know who wrote them, limit what they can access and use narrow credentials. The AI agent security checklist covers how, and prompt injection explains how text an agent reads can steer the tools it uses.

Sources

  1. Model Context Protocol: What is MCP?(modelcontextprotocol.io)
  2. Anthropic: Introducing the Model Context Protocol(anthropic.com)