Have a way to "see" exactly what the string being passed to an MCP Action Node or an API Call Node as created by Agentic Flow...

I have spent much time messing around with the parameters fields in these nodes to try to get exactly the right “sauce” that the server wants so as to actually accept the call and then be able to act upon it in a meaningful way.

It would be great if there were some mechanism where we could see EXACTLY what the string(s) are that are being passed to the server so that we could tweak our input parameters, etc., to help the system to form the request/commands in a way that works.

An example of this that made me think of it, was that last week I was attempting to connect my install of NanoBrowser to a local instance of LMStudio for an LLM connection. Everything I tried failed, even when attempting to connect to other supposedly OPENAI API compatible providers, Ollama, or cloud, Abacus. Nothing worked and I was pulling my hair out.

By chance I had LMStudio UI open on the one machine with the developer screen open which has a log screen as calls are made to its models. I happened to glance at the log screen just as I was making yet another attempt at a call, an I saw the issue: Nanobrowser had a “bug” in it where it was injecting an extra piece of the endpoint URL into my entered configuration screen. So for example, if the endpoint should be: http://localhost:8000/v1/chatcompletions, it was sending http://localhost:8000/v1/v1/chatcompletions, thus breaking any string, while correctly entered by the user, and breaking the integration.

If I did not have the visibility, I would have NEVER known how or where to look or what the issue actually was…

So, it would be cool to have that functionality in those nodes in Agentic Flow.

Cheers!