MCP Servers times out

[AhmedRaafat14]

I've been running Flowise on my local machine for some time now, and everything works as expected.

Once I have deployed it on an AWS instance, when I add an Agent node, then add a Custom MCP to it, like this one, for example:

{
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-everything"
      ]
    }

Flowise starts to throw error: 'MCP error -32001: Request timed out' This is a bit weird, especially since I don't have that issue locally and never did!

I thought it might be the internet, but I can confirm that my ECS instance has access to the internet to download the server (npm package). Also, thought maybe it was permissions issues, but can't debug that fully, and it actually worked for some time on AWS, then broke again, which made me think it is not really a permissions issue.

Does anyone have any idea why this might be happening?

[HenryHengZJ]

npx is really meant for local use, not when deployed to cloud. when it is deployed to AWS, its recommended to use streamable HTTP: https://docs.flowiseai.com/tutorials/tools-and-mcp#custom-mcp

[AhmedRaafat14]

npx is really meant for local use, not when deployed to cloud. when it is deployed to AWS, its recommended to use streamable HTTP: https://docs.flowiseai.com/tutorials/tools-and-mcp#custom-mcp

Thanks @HenryHengZJ for the reply, that is interesting, but how do you trigger the streambleHTTP MCP server in Flowise?

If i use the same example server from above:

{
  "command": "npx",
  "args": [
    "-y",
    "--",
    "@modelcontextprotocol/server-everything",
    "streamableHttp"
  ]
}

This is how to run that server in StreambleHttp, but I can see in the logs it still builds a Stdio object and tries to run it over stdio!

Do I have to run it first, then have a URL to be able to use it? Can't it just install and start it!

[HenryHengZJ]

here's more info: https://docs.flowiseai.com/tutorials/tools-and-mcp#streamable-http-recommended