Mendix MCP Server Thread

[XiaoHao-zxh]

When integrating MCP Server with Mendix 10.24.3, I encountered an issue. After AI initiates a request using the SSE protocol, a thread is created in the McpServerRequestHandler of the JAVA Action. Each time AI calls the tool, a new thread is created. Then I found a section of code in McpServerRequestHandler:

AsyncContext asyncContext = request.startAsync();

asyncContext.setTimeout(15 * 60 * 1000);

Thread.sleep(15 * 60 * 1000);

I want to know the significance of the line "Thread.sleep(15 * 60 * 1000);" and the considerations behind adding this line of code. Based on my current understanding, when there are a large number of AI requests, this line of code will cause Mendix to have OOM (Out of Memory) due to a sudden surge in threads.
Our project has crashed multiple times because of this issue, and we would like to seek a solution. Thank you!

[liamsommer-mx]

Hi,

the line was added because prior to Mendix 10.24, connections were closed immediately after the request handler was finished, so we added a workaround. It is currently under investigation if we can remove it, because a fix for this issue was implemented into the Mendix runtime with 10.24.

We haven't noticed OOM issues yet (neither on our side or from other customers), so interesting to learn about this. Could you elaborate how your project crashed due to this?