Beyond the Agent Trace: Operating Foundry Agents with .NET Aspire

Building an agent is only the beginning. When a request produces a slow, incorrect, or incomplete result, developers need visibility into both the agent's execution and the application surrounding it. In this session, I'll share practical patterns from a procurement workflow built in C# that calls a Microsoft Foundry agent, using .NET Aspire for application observability and transactional business audit records. We'll compare agent-centric traces in Foundry with application-centric traces in Aspire, following a sourcing request through an API, supplier lookup, agent invocation, output validation, and persistence. Then we'll break the workflow: the agent succeeds, but the business transaction fails. Through this example, we'll explore trace correlation, business-stage instrumentation, dependency deadlines and retries, sensitive-data boundaries, and why telemetry and business auditing answer different questions.

Speakers

Shweta Iyer