Halo is an open-source tool that creates tamper-evident runtime evidence for AI agents, helping companies verify agent behavior for compliance. The project launches on Hacker News, addressing the need for transparency and auditability in post-agentic enterprise environments.