I am making some sort of customer service MCP server for my own use. I can tell an LLM (such as LFM2.5-2.6B-Q4_K_M.gguf
in llama-cpp
) what I want, and it will do it for me. I have been thinking for ways for an AI agent to convey to the MCP server what the user is talking about.
Basically, if the user says something like "view tutorial XYZ", the AI agent needs to provide a hint
about what the person is asking about. The view_file
tool has an optional hint
parameter, and the tool description says to use hint
"t" when the user is talking about tutorials.
Then in the Rust MCP server, the code can use the hint
(alongside the name
) and use it to help find the appropriate page. So if the user is talking about tutorials, it will return a page that more closely matches tutorials rather than blog posts.
The goal of the LLM is to parse in natural language and provide my own functions (like view_file
) with information about what the user was talking about. The hint
can give better results than just a name
; having the LLM help determine the context of the request is useful.