What the demo covers:
The Ensage AI dashboard, showing sanctioned and unsanctioned (shadow) agents by endpoint and user, with usage trends and session logs
A sandboxed Claude agent limited by policy to Anthropic endpoints. Its requests to an outside website and to an internal Gitea server are blocked.
One policy change that grants Claude access to the Gitea server
Adding OpenAI endpoints to a policy so Codex, blocked at the start, can run
Each agent gets access to the specific resources it needs and nothing else, and you can extend access by editing policy from the console.
