Issue with experimental_createMCPClient hanging when connecting to remote MCP server deployed as Azure Function

[Praneethtkonda]

Hi team,

I'm currently working on integrating a remote MCP server (deployed as an Azure Function) with the @ai-sdk/experimental_createMCPClient method. I'm using the vercel ai SDK with Remix framework and have followed the official Azure Functions MCP example here:
https://github.com/Azure-Samples/remote-mcp-functions-typescript/tree/main

However, when I try to connect to the locally hosted Azure Functions MCP server via SSE using the following config:

const azureFunctionsMCPClient = await experimental_createMCPClient({
  transport: {
    type: 'sse',
    url: 'http://localhost:7071/runtime/webhooks/mcp/sse',
  },
});

The code hangs indefinitely at the experimental_createMCPClient call and doesn’t proceed or throw any error. Other transports (like Experimental_StdioMCPTransport and local SSE endpoints) are working fine in the same setup.

I've confirmed that the Azure Functions MCP server is running and reachable at the specified URL (I can hit it via the browser and get an SSE response). Still, the SDK doesn't seem to progress past the experimental_createMCPClient instantiation.

Here are a few things I’ve tried:

Verified CORS and headers on the Azure Function.

Confirmed the endpoint returns valid SSE.

Ran with and without tools registered in the function.

Would really appreciate any guidance on:

Whether this is a known issue with experimental_createMCPClient and remote Azure Functions.

Any debug strategies or logging support to better inspect what's happening internally.

Compatibility status of Azure Functions + MCP server with the vercel ai SDK.

Thanks in advance for your help!

[twxia]

It might be the parser issue I have patched at #6197. Please upgrade your ai package.

[Praneethtkonda]

@twxia I have updated to the latest version of ai sdk i.e 4.3.15, but still facing the same issue.

But if I do this way, it is working this way with the experimental_createMCPClient :-

import { SSEClientTransport } from '@modelcontextprotocol/sdk/client/sse.js';

const azureFunctionsTransport = new SSEClientTransport(
  new URL('http://localhost:7071/runtime/webhooks/mcp/sse'),
);
const customAzureFunctionClient = await experimental_createMCPClient({
  transport: azureFunctionsTransport,
});

Here I have to use the official model context protocol sdk for the SSE transport.
Any reasons why the sdk is not supporting remote mcp azure functions ?