A unified Azure platform for creating and managing AI models, agents, and applications with built‑in enterprise security, monitoring, and governance
Hello @Glen Anderson and @Kienan McLellan
Based on the additional testing in this thread, this now looks like a service-side regression affecting the Anthropic API integration in Microsoft Foundry, rather than an issue specific to your resource or Sweden Central.
The strongest indicators are:
- /anthropic/v1/files previously worked and began returning 404.
- The same behavior has now been reproduced in East US 2.
- /anthropic/v1/skills has also started failing.
- Existing skill IDs can still be used, suggesting the underlying model deployment itself remains functional.
Microsoft’s current Foundry documentation explicitly lists both the Files API and Skills API as supported Claude API surfaces, so a 404 Not Found from these routes shouldn't be treated as expected behavior.
Since the issue has now appeared across multiple regions and tenants, I wouldn't recommend recreating deployments, changing authentication, or modifying endpoint URLs as a workaround.
Please preserve the request-id, apim-request-id, UTC timestamp, region, and affected endpoint from each failure.
At this point, I recommend that a Microsoft moderator help escalate this to the Microsoft Foundry/Anthropic engineering team to check for a routing or API gateway regression affecting the Files and Skills endpoints.
The fact that East US 2 was working earlier and subsequently began returning the same errors should be particularly useful for engineering when correlating the start of the incident.
Please "Accept the Answer" if this information helped you. This will help us and others in the community as well.