Model Context Protocol (MCP) is an integration standard that lets AI applications connect to external servers exposing tools, resources, and prompts. In practice, it is a bridge between a model and the systems it can query or control, such as files, databases, APIs, or automation tools.
MCP matters in cyber security because that bridge is a trust boundary. If an attacker compromises an MCP server, or tricks a client into using a malicious one, they may gain access to sensitive data, credentials, or privileged actions. Defenders should treat MCP endpoints like any other high-value service: limit exposure, enforce strong authentication and authorization, sandbox tools, log requests, and review what each server can actually do. In attacks, MCP-like infrastructure can be scanned for weaknesses, abused to collect secrets, or used to pivot into broader systems; in defense, careful isolation keeps AI integrations useful without becoming an easy route to execution.