Title: MCP security: what breaks when you connect an untrusted tool server
The Model Context Protocol makes it trivial to plug tools into an agent -- including tools you don't control. Here are the real risks and how to contain them.
Read more →Blog
Notes on agent security and MCP
Updates on Enforgate, and deeper looks at the action-boundary model behind it.
Default-deny is the only safe posture for AI agents
Allowlists beat blocklists when the actor is non-deterministic. Here's why Enforgate denies by default at the action boundary -- and what that means in practice.
Read more →Securing AI agents: how Enforgate addresses the OWASP Agentic Top 10
AI agents that can call real tools need a real security boundary. Here's what Enforgate is, how it stops agents from doing the wrong thing, and how it maps to the OWASP Agentic Security Top 10.
Read more →