Python: [Bug]: Unable to access /responses metadata field in Agent Framework Workflows

[ashishmundra4001]

Description

Issue: How to access HTTP Request Information in Hosted Workflow Agents
I have a Foundry Hosted Agent. I created a Executor. Code in code sample section below.

I call this Agent over /responses api like below:

{
"agent": {
"name": "Super-Agent",
"type": "agent_reference"
},
"stream": false,
"input": "Which phone from Google is allowed in University? Answer using tools that you have available.",
"metadata": {
"x-some-token": "some token here"
}
}

Related Questions:

1. How do I read this metadata input parameter. When I had basic Agent without a Workflow, I could get the metadata in AgentMiddleware like below, but now that I am using Workflow, how do I get this metadata in my Workflow Executor?

2. Is AgentMiddleware not supported when using Workflows? I see FunctionMiddleware gets called, but FunctionMiddleware kwargs was empty, probably as I populate them in AgentMiddleware and AgentMiddleware is not invoked.

class UserContextAgentMiddleware(AgentMiddleware):
"""Agent middleware that obtains user context."""

async def process(
    self,
    context: AgentContext,
    next: Callable[[], Awaitable[None]],
) -> None:
    if hasattr(context.agent, '_request_headers'):
        request_data = context.agent._request_headers

        # Runtime values that will be forwarded to tools
        if 'x-some-token' in request_data:
            context.kwargs.setdefault("some_token", request_data['x-some-token'])
            context.metadata["some_token"] = request_data['x-some-token']

Code Sample

Run Workflow as HostedAgent


        agent_executor = AgentExecutorWithMCPTools(config)
        
        # Build workflow with connected executors
        workflow = (
            WorkflowBuilder(start_executor=agent_executor)
            .build()
        )

        # Run the agent - this handles its own event loop
        from_agent_framework(workflow.as_agent()).run()


Workflow Executor



class AgentExecutorWithMCPTools(Executor):
    """
    Workflow executor that runs an agent with the initialized MCP tools.
    
    This executor receives the MCP tools from the initialization node
    and creates an agent with those tools to process the user's request.
    """
    
    def __init__(self, config: AppConfig, id: str = "agent_with_mcp_tools"):
        super().__init__(id=id)
        self.token_service = None
        self.config = config
    
    @handler
    async def run_agent(
        self, 
        user_input: List[Message], 
        ctx: WorkflowContext[Dict, str]
    ) -> None:

            async with mcp_tool:  # Ensure proper cleanup
                # Create agent with the initialized MCP tool
                agent = chat_client.as_agent(
                    name="SomeAgent",
                    instructions=instructions,
                    tools=[mcp_tool],
                    middleware=[UserContextAgentMiddleware(), UserContextFunctionMiddleware()]
                )
                
                logger.info("Running agent with user query...")
                
                # Run agent with user input
                response = await agent.run(user_input)
                
                logger.info("Agent completed successfully")

Error Messages / Stack Traces

Package Versions

azure-ai-agentserver-agentframework==1.0.0b17 agent-framework

Python Version

3.12

Additional Context

No response