I Gave 4 AI Agents a Corporate Bank Account. Here’s How I Stopped Them From Draining It.

**Executive summary:** This build log describes Project ClawControl, a demo multi-agent finance control room that enforces identity-based governance by routing all agent API calls through an AI Identity Gateway (Maverics), evaluating OPA Rego policies per-tool-call, performing RFC 8693 delegation exchanges to fuse human and agent identity, issuing 5-second scoped tokens, and providing audit logging, escalation messaging, and container hardening; it is an implementation and design discussion rather than a security incident report.