shantoislamdev/claude-adapter — reverse-engineered prompt
Reverse engineered prompt
Build me a small TypeScript tool that lets Claude Code talk to any OpenAI compatible model through a local proxy. I want it to feel like Claude Code is using Anthropic normally, but behind the scenes your app should translate the requests and responses so I can point it at OpenAI, DeepSeek, Grok, or a local compatible endpoint.
Make it run as a CLI with a simple first time setup wizard that asks for the base URL, API key, and which upstream models should stand in for opus, sonnet, and haiku. After setup, it should start a local server on a configurable port, default around 3080, and optionally update Claude Code settings for me.
Please make sure normal chat works, system prompts are preserved, streaming responses still work in real time, tool calling is translated both ways, and conversation history carries over correctly. If it makes sense, also expose a basic programmatic API for starting the server and using the request and response conversion functions directly. Look up current docs online if you need to.
Want more depth? Deep Reverse