Hello folks, 👋 I'm Ani, founder of Delphi. I spent 13 years in enterprise security before building this. Here's what pushed me to start: AI agents have stopped just answering questions. They now take real actions, moving money, accessing data, calling tools, talking to other agents. And the moment an agent can act, it becomes a target. Attackers can hide instructions inside normal-looking text and quietly hijack what the agent does. The tools most teams rely on can't see this. A firewall or a log tells you what happened after the fact. We wanted something that catches the attack at the moment of action, inside the agent, before it executes. So we built Delphi: runtime threat detection and governance for AI agents. It runs in the execution path, blocks malicious actions in real time, gives you one view of every agent across your fleet, lets you write your own policies, and generates the evidence you need for compliance. Fast enough to sit inline (under 100ms), and accurate (98% in our testing). We're early and building in the open, and I'd genuinely love your feedback, especially if you're deploying agents and thinking about how to secure them. What would you want a tool like this to do? Happy to answer anything.

Delphi Security

Runtime Security And Governance For AI Agents

Runtime Security And Governance For AI Agents