You'd reach for this when you need to build semantic understanding of a codebase beyond what grep or simple text search can offer. It indexes your code with AI-powered analysis, letting you query by concept rather than exact string matches. The knowledge graph component maps relationships between functions, classes, and modules, which is handy when you're trying to understand dependencies or trace how data flows through unfamiliar code. Think of it as turning your repository into a queryable knowledge base where you can ask "what handles user authentication" instead of hunting for specific function names.