Plan and manage an Azure AI solution 1.1 Choose the appropriate Foundry services for generative AI and agents
An agent in Foundry Agent Service returns a 'tool not supported' error for code…
Study guide skill: Choose appropriate memory, tool, and knowledge integration services for agent solutions
Question B6-004
An agent in Foundry Agent Service returns a 'tool not supported' error for code interpreter, even though the model support table shows that its model supports code interpreter. The model is deployed in the project. What should you check first?
Select one answer.
Answer: D
Tool availability requires support from both the model and the region, so if the region table shows No, the tool can't run even when the model table shows Yes. Setting tool_choice to required addresses an agent that doesn't call a tool, and better descriptions address empty or irrelevant results. Endpoint reachability checks apply to MCP and OpenAPI tools, not to the built-in code interpreter.
Deep dive
How it works
In Foundry Agent Service, a tool runs only when both the model and the project's region support it. A model that supports a tool in general might not support it in every region. Microsoft's own example: code interpreter doesn't run in regions that show no for Code Interpreter (such as southcentralus and spaincentral), regardless of which model you use.
The tool best practices page lists this exact symptom in its troubleshooting section: Foundry returns a "tool not supported" error even though the tables show support, because tool availability needs both the model and the region. The fix:
- Check the region availability table for your region, as well as the model support table.
- Confirm the model is actually deployed in the project and region you're targeting.
- Try a different region, or a different model deployment that supports the tool.
The model is already confirmed and deployed, so the region table is the first thing to check.
The other options, and when you'd use them
- Check the MCP endpoint: Endpoint reachability is the check for MCP and OpenAPI tools. Code interpreter is built in.
- Rewrite the tool description: Better descriptions and instructions are the fix when tool calls return empty or irrelevant results, not for a "not supported" error.
- tool_choice required: That's for an agent that doesn't call a tool. Here the service rejects the tool itself.
Exam tip
"Tool not supported" despite model support = check the region table. A tool needs model and region support.
References
Last reviewed: