More detailed error message

[EelcoKoster]

Is your feature request related to a problem? Please describe.
From my C# project I try to connect to a MCP server in Python.

I successfully connected from C# to a streaming-http mcp server. So the code works and looks like this:

try
{
    var mcpClient = await McpClientFactory.CreateAsync(new SseClientTransport(new()
    {
        Endpoint = new Uri("http://localhost:7071/mcp/"),
        TransportMode = HttpTransportMode.StreamableHttp,
    }, factory), new()
    {
        ProtocolVersion = "2025-06-18",
        ClientInfo = new() { Name = "AI Test client", Version = "1.0" }
    },
    factory);
    var tools = await mcpClient.ListToolsAsync();
    kernel.Plugins.AddFromFunctions("Hello", tools.Select(t => t.AsKernelFunction()));
}
catch (Exception e)
{
    Console.WriteLine($"ERROR Connecting to MCP server: {e.Message}");
}

I created a python MCP server. And tested it with MCP Inspector and Postman succesfully.
But when I try to connect to the MCP server with the above C# code. I get the error message "Response status code does not indicate success: 400 (Bad Request)". This is not very helpfull. There is probably a response text with the reason for the bad request. But I can not see it.

Errorlog is:
[2025-09-24T14:01:17.309Z] fail: ModelContextProtocol.Client.McpClient[1155727496]
[2025-09-24T14:01:17.313Z] http://localhost:7071/mcp/ client initialization error.
[2025-09-24T14:01:17.325Z] System.Net.Http.HttpRequestException: Response status code does not indicate success: 400 (Bad Request).
[2025-09-24T14:01:17.327Z] at System.Net.Http.HttpResponseMessage.EnsureSuccessStatusCode()
[2025-09-24T14:01:17.329Z] at ModelContextProtocol.Client.StreamableHttpClientSessionTransport.SendMessageAsync(JsonRpcMessage message, CancellationToken cancellationToken)
[2025-09-24T14:01:17.330Z] at ModelContextProtocol.McpSession.SendRequestAsync(JsonRpcRequest request, CancellationToken cancellationToken)
[2025-09-24T14:01:17.334Z] at ModelContextProtocol.McpEndpointExtensions.SendRequestAsync[TParameters,TResult](IMcpEndpoint endpoint, String method, TParameters parameters, JsonTypeInfo1 parametersTypeInfo, JsonTypeInfo1 resultTypeInfo, RequestId requestId, CancellationToken cancellationToken)
[2025-09-24T14:01:17.343Z] at ModelContextProtocol.Client.McpClient.ConnectAsync(CancellationToken cancellationToken)
[2025-09-24T14:01:17.347Z] info: ModelContextProtocol.Client.McpClient[2081236114]
[2025-09-24T14:01:17.350Z] http://localhost:7071/mcp/ shutting down.
[2025-09-24T14:01:17.351Z] info: ModelContextProtocol.Client.McpClient[2103773711]
[2025-09-24T14:01:17.353Z] http://localhost:7071/mcp/ message processing canceled.
[2025-09-24T14:01:17.354Z] info: ModelContextProtocol.Client.McpClient[613617602]
[2025-09-24T14:01:17.356Z] http://localhost:7071/mcp/ shut down.

Describe the solution you'd like
When an error happens. I would like to also see the http response text so I can find the reason of the faulty http status.

Extra context
Python code (simplified) of MCP server (this is a functionapp with FastAPI and timertrigger + MCP):

import azure.functions as func
from fastapi import FastAPI
from mcp.server.fastmcp import FastMCP

mcp = FastMCP("Hello MCP", streamable_http_path = "/mcp/", json_response=True)

@mcp.tool()
async def hello(name: str) -> str:
    """Says hello to the user."""
    logging.info(f"Hello function called with name: {name}")
    return f"Hello, {name}!"

fast_app = FastAPI(title = "Hello API")

@fast_app.get("/hello")
async def root():
    return {"message": "Hello World"}

mcp_app = mcp.streamable_http_app()
mcp_app.mount("/", app=fast_app)

app = func.AsgiFunctionApp(app=mcp_app, http_auth_level=func.AuthLevel.ANONYMOUS)

@app.timer_trigger(arg_name="timer", schedule="0 0 8 * * *")
async def hello_timer(timer: func.TimerRequest):
    logging.info(f"Hello timer triggered!")

[mehrandvd]

What error message would you expect to receive? If you create your MCP client in Python, what error message will appear? I assume no additional information will be displayed in that scenario either.

[EelcoKoster]

Hi Mehran, thanks for your reaction.

The problem is that with other MCP clients like MCP Inspector and Postman I do not get an error. It just works. Only the C# gives an error.

But if I make an intentional error with Postman. Like sending a Content-Type: "application/j", the MCP returns statuscode 400 Bad request. But in the body of the response is "Invalid Content-Type header". The body text is essensial to understand the reason.

The main reason for this issue is that only this code returns an error and I'm struggling to find the reason. Not only returning the status code but also the body text can help me (and maybe some other developers) to understand why a 400 Bad request occurres.

[EelcoKoster]

I finally discovered what’s causing the issue. With a lot of trail and error I found that the McpClientFactory is using Transfer-Encoding=Chunked while initializing the connection. (MCP Inspector and my Postman tests do not).

The error for the 400 Bad request was: "Transfer-Encoding: chunked' header can not be used when content object is not specified"
This is pretty weird because the content object was actually filled.

Turns out that running Azure functions locally doesn't support chunking!! Deployed the code to an actual Azure Function and there it worked properly!

I still think that adding the body text to the error message would help developers in these edge cases...

Another consideration: Make the chunked transfer encoding optional? Not all servers handle this well -- I've seen more issues here with hanging connections. I trust the .NET implementation. But it takes two to tango :-)