fkiene/llmtrim ? reverse-engineered prompt

Build me a local app called llmtrim that sits between my coding tools and LLM providers and makes the requests smaller so I pay less, but the answers stay the same.

I want it to work with tools like Claude Code, Cursor, Codex, and anything else that can use a local HTTPS proxy. It should trim obvious waste from prompts, chat history, tool output, and repeated code before sending it out, then pass the reply back unchanged. Please include a simple setup flow, a status screen or command that shows savings and whether everything is healthy, plus commands to start, stop, diagnose, update, and uninstall it.

If it makes sense, also add a small tray app and a way to use it as an MCP server or library. Keep everything local, safe, and easy to install on macOS, Linux, and Windows. Look up current docs online if you need to.