Extracting JSON schema's from tools when used as an MCP client

[lmorg]

Hi, I'm using this package as an MCP client (tied into langchaingo to manage the LLM as an agent).

One problem I have is that the LLM isn't aware of how to pass parameters to MCP, which I believe is because I'm not exposing the schema correctly. But the question I have is, how do I discover that schema?

the MCP client shows the schema as a Go type, which is great for parsing the LLM tool input. But it's not so great for grabbing the schema and sharing with the LLM.

I've shared a little more background into my workflow here: #271 (comment) but essentially I've got the agent piece working really well now (via langchaingo). And I can create tools via langchaingo too. But I was looking for a way to integrate langchaingo with mcp-go. I have written a wrapper that kind of works. But the real problem is that the LLM doesn't really know how to pass parameters to an MCP via langchaingo. My thinking is the biggest hurdle here is that the schema isn't shared.