I built Pliuz because I watched an AI agent in production send 47 duplicate Stripe refunds at 03:14 in the morning — and the postmortem revealed nobody could actually answer why it had been allowed to act at all. Every team I talked to since had a version of that story.
The pattern was always the same: someone shipped an agent, the agent worked great for a week, then it took an action somebody had to apologize for. Approval handlers got hand-rolled per agent. Audit logs lived in Slack history. CISOs found out during pen-tests, not before. The cost of the first incident is what makes everyone serious — by then it is too late.
Pliuz is the layer I wished existed. One decorator wraps the risky call. JSONLogic policies auto-approve the boring stuff. A SHA-256 hash chain proves what happened to anyone who asks. EU-hosted, RLS-isolated, no LLM in the critical path. Open-source SDKs because you should be able to fork us if we get this wrong.
I am a solo technical founder. Pre-revenue. Building in the open. If you are running agents in production and Pliuz looks useful, talk to me directly — design-partner pricing is real and I will personally help you ship.