I built something in a similar space: Linko (https://github.com/monsterxx03/linko https://github.com/monsterxx03/linko), a transparent MITM proxy with a webui t

I built something in a similar space: Linko (https://github.com/monsterxx03/linko https://github.com/monsterxx03/linko), a transparent MITM proxy with a webui that lets you see what's actually being sent between Claude Code and LLM APIs in real time. It's been really helpful for me to debug my own sessions and understand what the model is seeing (system prompts, tool definitions, tracing tool calls etc.).