Support _meta in mcp_tool

[newhardwarefound]

Description
Currently, the @mcp_tool decorator only supports static metadata defined at registration. However, the MCP Protocol supports a _meta object within the tools/call request parameters. This field is essential for passing dynamic, request-scoped context.

In the current Azure Functions implementation, this _meta information is dropped and is not accessible via func.MCPToolContext.

Observed Behavior
When a client sends a tool call with a _meta block:

{ "method": "tools/call", "params": { "name": "get_job_appointment_details", "arguments": {}, "_meta": { "progressToken": 1, "tenant_id": "123", "user_id": "466", "appointment_id": "789" } } }
And the function is defined as:

@mcp_job.function_name("mcp_get_job_appointment_details")
@mcp_job.mcp_tool()
def get_job_appointment_details(context: func.MCPToolContext):
    # Attempting to access metadata
    pass

The context object contains transport headers and basic tool info, but the _meta fields (like tenant_id or appointment_id) are missing entirely from the payload.

Expected Behavior
The MCPToolContext should be updated to include a meta property (or similar) that captures the _meta object from the MCP request. This would allow developers to perform multi-tenant lookups or telemetry tracking based on the caller's specific context.

[newhardwarefound]

@hallvictoria I checked the example in https://github.com/Azure-Samples/remote-mcp-functions-python/blob/main/src/function_app.py and don't find a good solution. We care about this much more than how to supply front end UI from back end.

[hallvictoria]

Hi @newhardwarefound, thanks for reporting. We're currently investigating this and will let keep this issue updated with our findings.

// transferring to the worker repo for better tracking