It is crucial to strike a balance between the challenges of educational games and the abilities of the learners as players, especially in educational games that involve problem solving, such as tile-based programming in computational thinking learning. Most of the current tile-based programming games only offer a tutorial feature that instructs players to complete the game step by step. While this feature can assist players struggling with challenges, it may also cause boredom for those who do not require help in completing them. This research proposes the application of an adaptivity approach through the development of a Non-Player Character (NPC) that acts as a pedagogical agent in a tile-based programming game. The pedagogical agent feature designed as a finite state machine and implemented by a rule-based system in this research has the advantage of providing assistance according to the desires and needs of the players.