The Model Context Protocol (MCP) is quickly becoming a standard for AI agents and servers, defining how agents discover, authenticate to and invoke remote tools and services. But securing OAuth-based MCP servers is trickier than it looks.
Recent updates to the MCP Security Best Practices specification, led by security experts including Den Delimarsky and Paul Carleton, have highlighted critical gaps in current deployments, particularly around confused deputy attacks and token-handling vulnerabilities. In this article, I’ll walk through how…