Andon is a pre-execution trust gate for AI agents. Before an agent hits an API, URL, or payment, Andon returns Approve / Block / Human Review in ~200ms—backed by an auditable evidence chain (no source, no output). Built for agentic payments, procurement, and tool-use workflows where “oops” is too late.