LL(k) grammar is a formal grammar used in computer science for parsing, where 'LL' stands for Left-to-right scanning of the input and Leftmost derivation, and 'k' specifies the number of lookahead tokens needed to make parsing decisions. It is used in constructing top-down parsers, which are efficient for certain classes of languages but can be limited by the need for a fixed number of lookahead tokens.