Older adults often encounter unfamiliar procedural tasks when utilizing mobile services. Existing solutions, such as beginner guides or instructional manuals, are less effective for this demographic since older adults prefer a “learn by doing” approach rather than “learn before doing”. This research aims to design a context adaptive tutorial agent that leverages state-of-the-art multimodal Large Language Models (LLMs) to address the adaptation and personalization problems in procedural smartphone tasks for older adults.