Network Error while working with Microsoft Foundry Workflows

Question

Screenshot_18-5-2026_232159_ai.azure.com.jpegNetwork Error while working with Microsoft Foundry Workflows

Using Microsoft Foundry Workflows (preview) and its ending up stating network error.

I am using Multiagent declared already in the Agents space. Workflows created by 5+ agents. Using Microsoft Foundry Workflows (preview) and its ending up stating network error.

Any guidance on troubleshooting.

Answer 1

Hi Amira,

Thank you for your response.

I wanted to share the current setup and challenges for guidance:

Current Approach

• I initially created all agents and tested them individually — everything worked as expected.
• I then created a Foundry Workflow and orchestrated these agents together.
• All agents are natural language generation-based, with no coding involved.

Initial Issue & Resolution

• When running the full workflow, I encountered a network error.
• Based on a colleague’s suggestion, I:
  • Optimized the prompts
    • Added instructions in all agents to limit token usage in outputs
    • This resolved the issue, and the workflow started functioning correctly.

New Challenge

I introduced a new agent for diagram/image generation using the Code Interpreter tool.

This agent:

• Works fine when tested independently
  • Has a significantly higher response time (~300+ seconds)
  After integrating this agent into the workflow (between existing agents), the network error started occurring again.

Constraint

• Since this is not a language-generation agent, prompt optimization is not very applicable here.

Ask

• I’m unsure how to handle this scenario, especially with the longer execution time of the Code Interpreter agent.
• Looking for guidance on:
• Best practices for integrating such high-latency agents into workflows
  • Any architectural or workflow design recommendations to avoid these errors

Answer 2

Hello Nitesh !

Thank you for posting on MS Learn Q&A.

Because Workflow agents are still in preview in Microsoft Foundry Agent Service, I would troubleshoot it in layers rather than assuming it is a pure browser network issue. Workflow agents are a preview agent type and workflows are designed to orchestrate multiple agents steps, so a failure in one child agent can appear as a generic runtime error in the UI.

Open every agent used in the workflow and test it directly from the Agents playground. If one agent fails alone, fix that agent before testing the workflow.

And since your workflow has 5+ agents, create a temporary workflow with only 1 or 2 agents. If that works, add agents back one by one. This helps identify whether the problem is caused by a specific agent, tool, model deployment or handoff step.

Open the browser DevTools then Network tab and rerun the workflow. Look for failed calls and capture:

• HTTP status code: 401, 403, 429, 500, 502, 504...
• request URL
• correlation/request ID
• response body if available

If your issue still persist, I suggest that you open a support ticket.