👋 Hi Product Hunt! AI coding agents are becoming increasingly autonomous. They can edit code, execute terminal commands, access files, and interact with external tools. While that unlocks incredible productivity, it also introduces a new challenge: trust. I built PRAXIS to be the runtime guardrail between an AI agent and execution. Instead of blindly allowing every action, PRAXIS evaluates each request in real time and decides whether to allow it, warn the user, require approval, or block it entirely. Every decision is recorded with a complete audit trail so developers know exactly what happened and why. Current capabilities include: • Runtime interception of AI tool calls • Risk classification • Policy-based approvals and blocking • Human-in-the-loop verification • Audit logging for every action This is just the beginning. The vision is to build a universal runtime safety layer that works across AI coding agents and developer tools. I'd love your feedback. • What AI coding tool do you use most? • What's the riskiest AI action you've seen? • What guardrails would make you trust AI agents more? Thanks for checking out PRAXIS!

PRAXIS

Stop risky AI actions before they execute

Stop risky AI actions before they execute