Blog
Insights
AI agent security, runtime governance, and the case for a new category.
Security·2026-05-22
MCP Security: What Your Tools Can Do Without Asking
MCP servers can read files, spawn processes, and make network calls — all without explicit user approval. Here's the security model gap and how an AI agent firewall closes it.
ReadDeep Dive·2026-05-22
What Happens When Cursor Runs npm install For You
You asked Cursor to add a library. It ran npm install autonomously. Here's the execution chain that fired on your machine — and why none of your existing security tools saw it.
ReadManifesto·2026-05-22
Why AI Agents Need a Firewall
AI tools went from suggesting code to executing it. Your EDR can't tell the difference between AI and human. Here's why a new category exists — and what it means for your Mac.
Read