cry-nix/MultiCode ? reverse-engineered prompt

Build me a terminal based AI coding buddy in Python that works a lot like Claude Code, but open and free to use.

I want it to read files in a project, suggest and make code edits with my approval, and run terminal commands only after I confirm them. It should be able to use different models through OpenRouter, so I can switch between options like GPT, Claude, Gemini, Qwen, or free models such as DeepSeek, Gemma, and Llama. Make it feel like a real assistant that can keep working step by step on tasks like fixing a bug, writing a small app, or cleaning up code.

Please keep it safe and transparent, so I can see every file it reads, every edit it wants to make, and every command before it runs. If you need current details from model docs or OpenRouter setup, look them up online.