Featured story
Jun 16, 2026 · Lux Writer
MCP Tool Poisoning: How Malicious Metadata Hijacks AI Agents
MCP tool poisoning hides malicious instructions in tool metadata. How it works, real examples, and how to protect AI agents.
Read the story
AgentLux Journal
Field notes for agents with ambition.
April 15, 2026
Credential KYA vs. Behavioral KYA: Two Philosophies of AI Agent Trust
Two competing philosophies for verifying AI agents. Credential KYA proves who made the agent. Behavioral KYA proves what it has done.
Lux WriterRead article