[CONSOLIDATED] A lot of `get_azure_resource_and_app_health_status` prompts do not trigger the corresponding tool as expected

[Menghua1]

Description:
In consolidated mode, when testing prompts related to the get_azure_resource_and_app_health_status tool, most prompts will pick up Azure Cli tool over the MCP tool, as shown below:

Steps to Reproduce:

1. Start mcp server in consolidated mode.
2. Open GitHub Copilot in VS Code and switch to Agent mode.
3. Click Refresh on the tools list, and select the Built-in, Extension: Github Copilot for Azure, MCP Server: Microsoft Docs, github MCP server and MCP Server: Azure MCP.
4. Enter test prompts Query an agent in my AI foundry project, Query and evaluate an agent in my AI Foundry project for task_adherence and Evaluate the full query and response I got from my agent for task_adherence.

Environment:

• OS: Windows.
• Vscode Version: 1.104.2
• GitHub Copilot Chat Version: 0.31.3
• Large Language Model: Claude Sonnet 4.5
• Virtual Tool: On
• Azure MCP Server: 0.9.7
• Test Prompts:

1. Get metric definitions for app service mhwebapp from the namespace
2. Analyze the performance trends and response times for Application Insights resource <resource_name> over the last <time_period>
3. Get the average CpuPercentage metric for app service 'mhwebapp' in group 'mhmcp' over the last 48 hour with intervals
4. Get the availability status for resource mhwebapp in group mhmcp
5. Investigate error rates and failed requests for Application Insights resource mhinsight in group mhmcp for the last 48h
6. List active service health events in my subscription
7. List all available table types in the Log Analytics workspace <workspace_name>
8. List all Azure Managed Grafana in one subscription
9. List all Log Analytics workspaces in my subscription
10. List all monitored resources in the Datadog resource <resource_name>
11. List all service health events in my subscription
12. List all Web Test resources in my subscription
13. List profiler recommendations for Application Insights in resource group <resource_group_name>
14. List the activity logs of the last month for <resource_name>
15. Show me my Log Analytics workspaces.

Expected Behavior:
These prompts can successfully trigger the expected mcp tool.

@joshfree for notification

[joshfree]

@fanyang-mono PTAL - setting this to November

[fanyang-mono]

@Menghua1 I picked a couple of the testing prompts listed here. LLM was able to invoke get_azure_resource_and_app_health_status fairly consistently.

I wonder if you start a new chat window for each test prompt? If not, I would suggest doing that from now on.