Keep codebase-memory indexed at session start, then query its code graph from a shell.
Answer code questions from the codebase-memory graph instead of grepping and reading files one by one. Use before grep, rg, find, or opening several files to understand code: "where is X", "who calls X", "what breaks if I change X", how modules fit together, data flow, dead code, or a symbol's source. Indexes the repo first when it isn't indexed yet.
