Share via

Microsoft Foundry (new) Error

GNHD_TKR 40 Reputation points
2026-02-24T02:10:27.3733333+00:00

Region : Japan

using Microsoft Foundry (new)

I've checked all the access control roles in the Azure portal. Is this perhaps a temporary outage?

Today, selecting an agent from Microsoft Foundry produces the error shown in the image below.

There were no issues until three days agoUser's image

Additionally, when attempting to create a new agent, the following error is displayed.

User's image

Since the agent cannot be opened, sending a chat to the model will result in an errorUser's image

Foundry Tools
Foundry Tools

Formerly known as Azure AI Services or Azure Cognitive Services is a unified collection of prebuilt AI capabilities within the Microsoft Foundry platform

0 comments No comments

Answer accepted by question author

SRILAKSHMI C 19,100 Reputation points Microsoft External Staff Moderator
2026-03-18T12:04:16.27+00:00

Hi GNHD_TKR,

Thank you for the update this is helpful, and I’m glad to hear the agent is accessible again.

Based on your description, this behavior is most consistent with a temporary service-side or portal-related issue, rather than anything in your setup.

What likely caused the issue

From the symptoms you shared:

“Agent not found”

403 / Forbidden errors when creating agents

ERR_BAD_REQUEST during chat

Then automatic recovery without changes

This typically indicates a transient inconsistency in the Microsoft Foundry service layer, such as:

Temporary backend service disruption or deployment

Authorization/cache desynchronization (RBAC or token validation delays)

Metadata/indexing delay where the agent exists but isn’t resolvable

Regional service instability (in your case, Japan region)

During such windows:

  • The portal may fail to resolve agents → “Agent not found”
  • API calls may fail authorization checks → 403 / Forbidden
  • Chat requests fail because the agent configuration cannot be retrieved

Why it started working again

Once the backend issue was resolved:

Agent metadata became available again

Access validation (RBAC/token checks) started succeeding

The portal was able to correctly load and interact with the agent

This explains why everything worked again without any changes on your side.

Is this expected?

This is not expected behavior under normal conditions, but it can happen occasionally due to:

  • Ongoing platform updates (especially in the new Microsoft Foundry experience)
  • Temporary service degradation not always visible in public status pages.

Thank you!

Was this answer helpful?

1 person found this answer helpful.
0 comments No comments

Answer accepted by question author

Jerald Felix 13,500 Reputation points Volunteer Moderator
2026-02-25T04:01:29.0733333+00:00

Hello GNHD_TKR,

Thanks for raising this question in Azure Q&A forum.

Based on your description selecting an agent from the new Microsoft Foundry portal suddenly producing an error that wasn't there three days ago this is a well-recognized pattern of issues affecting the new Foundry portal experience that multiple users have reported recently.

The most common causes for agents suddenly throwing errors in the new Foundry portal after previously working are:

1. Permission/RBAC issue after portal update The new Microsoft Foundry portal sometimes resets or fails to carry over role assignments when agents are re-loaded or when the portal itself receives backend updates. Go to Azure Portal → your AI Foundry resource → Access Control (IAM) and verify the following roles are still in place:

  • Azure AI Developer role on the project for the identity running the agent

Cognitive Services OpenAI User on the connected Azure OpenAI resource

  • If using tools (Bing, AI Search, etc.) verify the agent's managed identity still has access to those resources​

2. Connected resource or tool configuration broken Agents that use tools (vector stores, connected APIs, Logic Apps, etc.) can suddenly fail if the underlying resource has a token expiry, connection string change, or policy update. In the agent's Tools tab inside Foundry, review each connected tool and re-test the connection. Try removing all tools temporarily and testing with a minimal system prompt if that works, re-add tools one by one to isolate the culprit.

3. Model deployment quota or availability issue If the underlying model deployment (e.g., GPT-4o, GPT-4 Turbo) has hit its token-per-minute quota or is temporarily unavailable in your region, the agent will surface a generic error. Go to Models + Endpoints → Deployments in Foundry and check if the deployment shows as healthy. If it shows degraded or you're in Sweden Central, consider temporarily redeploying to East US or West Europe.

4. New Foundry portal known issue with agent selection There is a documented pattern where the new Foundry portal UI fails to load agent configurations properly due to a stale session or portal bug. Try:

Hard refresh (Ctrl + Shift + R) and clear browser cache

Switch from the new Foundry portal to the classic Foundry (ai.azure.com → use the toggle to switch to classic view)

Try a different browser or private/incognito mode

5. Re-deploy the agent If the error persists across browsers, the agent configuration itself may have become corrupted after a portal update. The quickest fix is to note down your agent's system prompt, model, and tool settings, delete the agent, and recreate it fresh. This resolves the majority of "worked before, broken now" cases in the new Foundry portal.​

To get the most targeted help, please share the exact error message or screenshot from the Foundry portal the error text or code will help pinpoint whether this is a permissions issue, model issue, or portal UI bug.

If it helps kindly accept the answer.

Best Regards,

Jerald Felix

Was this answer helpful?


0 additional answers

Sort by: Most helpful

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.