MashuPack lets developers select exact files or subsystems from local repositories and compile them into a single, structured text file formatted for large language models (notably ChatGPT and Claude). It operates entirely in the browser with no backend, account, or repo upload required, keeping code local and private while producing an LLM-friendly representation. The tool includes intelligent file merging that preserves structure and dependency context so models receive concise, coherent inputs instead of fragmented files. Its value is reducing context-window fragmentation and simplifying the process of feeding targeted, relevant code context to LLMs for review, summarization, or generation tasks.