Anthropic releases Model Context Protocol November 25 2024

This article compares the Model Context Protocol (MCP) and traditional APIs, explaining how MCP allows AI models to discover and invoke tools at runtime rather than relying on hardcoded developer calls. The author outlines the security implications of this shift, including risks like tool poisoning and over-scoped permissions, and describes how proxies like Formal's Connector can enforce least-privilege access and policy controls for MCP servers. Open article →