Many Domain Agents One Governance Model: An Azure Pattern | THR1001
Emilie Lundblad traces how domain-specific agents can share a governance model while retaining scoped identities and tool access. The session follows the architecture end to end within the customer’s own Azure tenant.
Architecture overview
When each business unit has its own agent, separate identities and tool permissions can create gaps in oversight. This production multi-domain architecture brings those agents under a common governance model through four components:
- Azure API Management AI gateway: Provides one governed entry point.
- Scoped Entra Agent IDs: Give each domain agent its own scoped identity.
- Governed MCP: Brokers tool access, with content safety checks applied to tool-call arguments.
- Foundry Control Plane: Provides oversight across the agent fleet.
The pattern combines domain-specific agents with shared controls at the entry point, identity, tool-access, and fleet-management layers. All of these operate within the customer’s own tenant.
Session details and resources
Presented at Microsoft Ignite 2026, this advanced, level-300 theater session is part of the Agent Platform track, with session code THR1001.