A framework to adjust sitting postures based on musculoskeletal models is presented in this paper. After the analysis of muscles and strength, some “comfortable” postures are specified as candidate postures. An optimization algorithm is then applied to choose from these candidates, according to current conditions of muscles and the durations one needs, and form a sequence of ordered postures for the sitters. The sitters shift the postures dynamically before they feel uncomfortable, therefore they can maintain a seated posture for extended durations. The ergonomists can also benefit from the information of intermittent postures and use it as a guideline for a new chair/desk design.