DataHubX — context-bound authorization for AI agents

The agent proposes. The policy decides. The human authorizes. The Gateway enforces — by re-reading DataHub itself immediately before the mutation and refusing if the world no longer matches what was approved.

Agent

The agent has three tools: list, inspect, propose. It has no tool that mutates DataHub.

Or propose directly

Govern

Every button drives the real domain against a real DataHub instance. Nothing here is simulated.

Action detail

action CHANGE_LIFECYCLEtarget customer_prodphase IDLE
Approved context
not captured yet
fingerprint
Current context
not captured yet
fingerprint

Audit timeline

No events yet — propose an action to begin.