FileScopeMCP provides semantic codebase awareness to AI assistants by continuously monitoring files, ranking them by importance based on dependency analysis, and maintaining AI-generated summaries through a background LLM broker. The server exposes tools including `find_important_files`, `get_file_summary`, `detect_cycles`, and `list_files` with importance scores, enabling Claude, Cursor, or other AI tools to understand code structure without reading source directly. It solves the problem of AI tools lacking real codebase context by automatically tracking dependencies across 12 languages, detecting changes via AST-level diffing, and serving a visual dashboard for exploring projects.
claude mcp add --transport stdio admica-filescopemcp uvx filescopemcp