MXC: Deploying Agents You Can Trust | BRK390
David Wiesen, Demetrius Nelon, and Stuart Schaefer present demonstrations of Microsoft Execution Containers (MXC), focusing on how Windows isolation choices apply to AI agent deployment.
Choosing an execution environment
The session emphasizes that a single sandbox approach does not fit every AI agent. It examines Windows isolation options and frames the choice of execution environment around three considerations:
- Workload requirements: What the agent needs to execute.
- Security considerations: The isolation requirements for that execution.
- Policy: The constraints that govern where and how the agent can run.
Microsoft Execution Containers provide the focus for demonstrations of these choices. The central takeaway is to assess the agent’s requirements rather than apply the same execution environment to every workload.
Session details and resources
- Event: Microsoft Ignite 2026
- Session: BRK390, Windows
- Format: Breakout
- Level: 400 — Expert
For additional Windows resources, see Windows next steps.