MCP - Cautions against granting wide access to AI agent

The security of AI agents isn't just about the model—it's about the tool permissions. MCP (Model Context Protocol) is great, but granting wide access to your filesystem or APIs is a recipe for disaster. Scope your tools, even in local dev. #MCP #AISecurity Source: https://x.com/natalie_avfieb/status/2033185150046245094