hoophq/hoop — reverse-engineered prompt

Build me an open source gateway that sits between our team and production systems so every database query, shell command, API call, Kubernetes action, or MCP request goes through one place first. I want it to mask sensitive data in responses before it reaches people or AI agents, block clearly dangerous actions like dropping tables or deleting things without a safe filter, and pause risky write actions for human approval in Slack or Teams. It should support real user sign in with our identity provider, keep full audit trails and session replay, and make it easy to see who did what and when.

The big goal is safe production access for engineers, support, service accounts, and coding agents without changing the tools they already use. It should feel fast, work as a gateway rather than an endpoint agent, and have a simple local setup with Docker so I can run it quickly. If you need details on protocols or auth, look up current docs online and wire up a clean web UI too.