Shared knowledge cache for AI coding agents — reuse an answer once it exists, before your model spends tokens.