Help needed: A2A client-server issues

[juan-saad]

Hi, I'm building a test application to understand A2A concept, for that, I've created an Azure Foundry project and deployed a gpt-4o-mini model to use as a chat client.
I've created both the server and client with the A2A as in the samples repository (using foundry agent with the Azure OpenAI endpoint) but whenever the client tries to execute the Agent as a tool, it fails with a 405 Http Status code while fetching the /v1/card using a POST method.
I don't understand why it's doing a POST instad of performing a GET.
Does someone had the same issue, or can guide me to troubleshoot it?

[moonbox3]

Hi @juan-saad are you using Python or .NET?

[juan-saad]

I'm using .NET.

Edit: I'm leaving some snippets of both client and server creation (both were created as WebApi projects) also consider this is being hosted using Aspire:

Server:

// filepath: Program.cs
IEndpointConventionBuilder a2aTaskManager = app.MapA2A(
    agent.Agent!,
    path: "/a2a",
    agentCard: agentCard,
    taskManager => app.MapWellKnownAgentCard(taskManager, "/a2a"));

// filepath: Agent.cs
// Custom class to hold different agent properties

public async Task InitializeAgentAsync()
{
    ChatClientAgent agent = new ChatClientBuilder(chatClient)
    .UseFunctionInvocation()
    .UseOpenTelemetry(
        sourceName: Name,
        configure: options =>
        {
            options.EnableSensitiveData = true;
        })
    .Build()
    .CreateAIAgent(
        name: Name,
        instructions: Instructions,
        tools:
        [
            AIFunctionFactory.Create(RandomNumber)
        ]);

    Agent = agent;
}

Client:

// filepath: Program.cs

app.MapPost("/chat", async (
    ChatRequestModel chatRequest,
    IChatClient chatClient,
    ILoggerFactory loggerFactory,
    IHttpClientFactory httpClientFactory,
    IConfiguration configuration,
    CancellationToken cancellationToken) =>
{
    ILogger logger = loggerFactory.CreateLogger("ChatEndpoint");
    HttpClient httpClient = httpClientFactory.CreateClient();

    var baseUri = configuration["A2A-SERVER-1_HTTPS"];

    AIAgent a2aAgent = await HelperClient.CreateAgentAsync(baseUri!, httpClient, loggerFactory);

    logger.LogInformation("HttpClient created with base url: {BaseUrl}", httpClient.BaseAddress);

    ChatClientAgent agent = chatClient
        .AsBuilder()
        .Use(getResponseFunc: CustomChatClientMiddleware, getStreamingResponseFunc: null)
        .Build()
        .CreateAIAgent(
            instructions: "You're a conversational AI Agent designed to help users in any task",
            name: "conversational-agent",
            description: "Very beautiful description for the agent",
            tools: [a2aAgent.AsAIFunction()]);

    var response = await agent.RunAsync(chatRequest.Message, cancellationToken: cancellationToken);

    return Results.Ok(response.Text);
});

// filepath: HelperClient.cs

internal static async Task<AIAgent> CreateAgentAsync(string agentUri, HttpClient httpClient, ILoggerFactory loggerFactory)
{
    var url = new Uri(agentUri);
    var logger = loggerFactory.CreateLogger("HelperClient");

    var agentCardResolver = new A2ACardResolver(url, httpClient, logger: logger);

    return await agentCardResolver.GetAIAgentAsync(httpClient, loggerFactory);
}