Microsoft Developer: When my agent is slow or wrong, where do I even look?

https://aka.ms/foundry-portal

https://aka.ms/InsideMicrosoftFoundryPlaylist



Your agent took too long to answer, or it answered with confidence and got it wrong. Was it the model, a tool, or the instructions? Without traces, you're guessing. With traces, you connect Application Insights to your Microsoft Foundry project, which turns on server-side tracing with no code changes. Then you read an agent run span by span: the agent invocation, each model call, and every tool call with its arguments, result, tokens, and latency. You find the slow step and the wrong step, then query the same data in Application Insights with KQL to see if it's one bad run or a pattern. By the end, you'll know exactly where to look first.



0:00 - Connect Agents to Enterprise Context

1:09 - Explore Web IQ

1:45 - Ground Responses with Foundry IQ

2:51 - Connect Data with Fabric IQ

3:57 - Query Semantic Models and Ontologies

4:33 - Take Actions with Work IQ

5:38 - Give the Agent Its Own Identity

6:11 - Automate Supply Escalations

7:18 - Manage Agents with Agent 365

8:22 - Add Teams Messaging Support