LLMCode

About

LLMCode demonstrates how much of the Claude Code workflow a compact terminal REPL can reproduce: gather context with /context or /#, request changes in natural language, and write the result with /write or /append, with workspace navigation and configuration handled by further commands. Because any OpenAI-compatible endpoint works, it ran against DeepSeek-R1 and local Llama 3 servers as readily as OpenAI itself, with settings stored in ~/.llm_code_config.json. The repository describes itself as under active development and partially built with its own assistance, but work stalled at 11 commits, leaving a prototype rather than a maintained tool.