[d4rkp4ttern]

Correct, no issues because since at least a few months, llama.cpp/server exposes an Anthropic messages API at v1/messages, in addition to the OpenAI-compatible API at v1/chat/completions. Claude Code uses the former.