Flexible planning allows for compromise in AI planning problems by associating preferences and satisfaction levels with goals and operators. This allows operators to relax preconditions at a cost to the overall plan satisfaction level. The document describes flexible planning graph (FGP) and leximin FGP (LFGP) approaches, which model the planning problem as a fuzzy constraint satisfaction problem. FGP searches for plans that maximize satisfaction level, while LFGP searches for plans that maximize satisfaction vectors lexicographically. Propagating satisfaction levels through the planning graph allows earlier pruning of search branches. LFGP is more computationally expensive but finds a broader range of high-quality compromise plans.