


Secure the tools an AI agent can reach
The MCP security guidance covers authentication and tool-access risks that become important when an assistant can take actions.
In this course
Inspect a practice MCP connection for authentication, server trust and excessive permissions, using only the authorised lab.
Run only in an authorised isolated lab, with controlled accounts and data.
Explore module 10Sources & expert videos
- MCP security best practices ↗Model Context Protocol · Documentation checked 19 November 2026








