What could be causing a deployed agent using MCP tool in Azure AI Foundry to get something went wrong error on azure bot service

Question

What could be causing a deployed agent using MCP tool in Azure AI Foundry to get something went wrong error on azure bot service

Fajra Risqulla 30

My agent is working well in playground, it can call/use mcp tool, but once I publish to microsoft teams and test it in bot service webchat, it return "somthing went wrong"

Fajra Risqulla 10 Reputation points

2025-12-02T19:27:59.42+00:00

Thank you, but I still got same issue, I have deployed in different regions. Previously, my agent is deployed in east us and then re-deployed it in sweden central. I've found nothing in bot service and Azure MCP Server logs. I really appreciate if you can help me to address this issue.
Arek Bar 5 Reputation points Microsoft Employee

2025-12-02T20:55:06.9933333+00:00

......
Brad C. Stevens 5 Reputation points Microsoft Employee

2025-12-02T22:07:56.65+00:00

Would love clarity on what regions and runtimes are limited vs. having to randomly test regions hoping one will work. :(
Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-03T13:14:37.7633333+00:00

**Hi Fajra Risqulla,
**Thanks for sharing the details! If the issue persists after redeploying the MCP agent to a different region, here are key steps to troubleshoot:

Validate Routing & Endpoints Ensure your Bot Service channels (Teams/Web) point to the correct MCP Server URL including /mcp route. Re‑save channel settings and re‑register the bot to clear stale endpoints.

Check Connectivity Test MCP health endpoint with curl to confirm DNS resolution and TLS reachability. Verify reverse proxies or App Gateways aren’t rewriting paths or dropping headers.

Authentication & RBAC Confirm Managed Identity or Service Principal is enabled on the new compute (App Service/Container Apps/AKS) and has the same RBAC permissions for Azure OpenAI, AI Search, Storage, and Key Vault.

Private Endpoints & Networking If using private endpoints, recreate them in the new region and update private DNS zones. Check NSGs, UDRs, and firewall rules for service tag access.

Dependencies & Attachments Detach vector store or knowledge base and test with minimal configuration. Reattach after baseline routing works.

Logging & Observability Enable Application Insights and Bot Service diagnostics. Add correlation IDs to trace requests and identify timeouts or upstream errors.

Minimal Repro Test Start with an echo tool (no model calls) and add dependencies one by one to isolate the failing component.

If these steps don’t resolve the issue, please share whether your deployment uses private endpoints or vector store attachments so we can suggest targeted checks.

Thankyou!

Answer accepted by question author

0 additional answers

Your answer

Fajra Risqulla 10 Reputation points

2025-12-02T19:27:59.42+00:00

Thank you, but I still got same issue, I have deployed in different regions. Previously, my agent is deployed in east us and then re-deployed it in sweden central. I've found nothing in bot service and Azure MCP Server logs. I really appreciate if you can help me to address this issue.
Arek Bar 5 Reputation points Microsoft Employee

2025-12-02T20:55:06.9933333+00:00

......
Brad C. Stevens 5 Reputation points Microsoft Employee

2025-12-02T22:07:56.65+00:00

Would love clarity on what regions and runtimes are limited vs. having to randomly test regions hoping one will work. :(
Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-03T13:14:37.7633333+00:00

**Hi Fajra Risqulla,
**Thanks for sharing the details! If the issue persists after redeploying the MCP agent to a different region, here are key steps to troubleshoot:

Validate Routing & Endpoints Ensure your Bot Service channels (Teams/Web) point to the correct MCP Server URL including /mcp route. Re‑save channel settings and re‑register the bot to clear stale endpoints.

Check Connectivity Test MCP health endpoint with curl to confirm DNS resolution and TLS reachability. Verify reverse proxies or App Gateways aren’t rewriting paths or dropping headers.

Authentication & RBAC Confirm Managed Identity or Service Principal is enabled on the new compute (App Service/Container Apps/AKS) and has the same RBAC permissions for Azure OpenAI, AI Search, Storage, and Key Vault.

Private Endpoints & Networking If using private endpoints, recreate them in the new region and update private DNS zones. Check NSGs, UDRs, and firewall rules for service tag access.

Dependencies & Attachments Detach vector store or knowledge base and test with minimal configuration. Reattach after baseline routing works.

Logging & Observability Enable Application Insights and Bot Service diagnostics. Add correlation IDs to trace requests and identify timeouts or upstream errors.

Minimal Repro Test Start with an echo tool (no model calls) and add dependencies one by one to isolate the failing component.

If these steps don’t resolve the issue, please share whether your deployment uses private endpoints or vector store attachments so we can suggest targeted checks.

Thankyou!

Answer 1

Hi Fajra Risqulla,

Thank you for reaching out on the Microsoft Q&A.
It seems like your agent’s MCP tool is working correctly in the playground but failing only after you deploy it through Azure AI Bot Service. This is a common pattern that others have reported as well.

One of the biggest causes appears to be region-related issues. Some users have found that MCP tools work in one Azure region but fail in another. For example, an agent that failed in West Europe worked immediately when deployed in Sweden Central. This suggests that MCP functionality may not be fully supported or consistent across all regions yet.

Another possibility is that the Bot Service runtime behaves differently from the playground. In several threads, users noted that the agent lists the tool correctly but never actually triggers a tool call once deployed. This can happen because of integration issues, restrictions in the hosted environment, or runtime differences with network or permissions.

To troubleshoot, you can try a few things:

Deploy your agent in a different Azure region to see if the issue disappears.
Enable detailed logging so you can confirm whether the tool call is even being attempted.
Test with a very simple MCP tool to check if the problem is with the tool or the environment.
Compare your playground settings with the bot’s runtime to make sure nothing differs in credentials or configuration.

This issue is likely not your code. It’s usually a region or runtime limitation with MCP in Azure Bot Service. Testing in another region is often the quickest way to confirm.

Please let me know if there are any remaining questions or additional details, I can help with, I’ll be glad to provide further clarification or guidance.

Thankyou!

Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-05T11:05:41.1133333+00:00

Hi Fajra Risqulla,

I’ve converted my earlier comment into an answer, could you please take a moment to mark it as Accepted? This helps others in the community with the same question find the solution more easily.

Share via

What could be causing a deployed agent using MCP tool in Azure AI Foundry to get something went wrong error on azure bot service

0 additional answers

Your answer