Incorporating soft actuation with soft yet durable textiles could effectively endow the latter with active and flexible shape morphing and motion like mollusks and plants. However, creating highly programmable and customizable soft robots based on textiles faces a longstanding design and manufacturing challenge. Here, we report a methodology of encoded sewing constraints for efficiently constructing three-dimensional (3D) soft textile robots through a simple 2D sewing process. By encoding heterogeneous stretching properties into three spatial seams of the sewed 3D textile shells, nonlinear inflation of the inner bladder can be guided to follow the predefined spatial shape and actuation sequence, for example, tendril-like shape morphing, tentacle-like sequential manipulation, and bioinspired locomotion only controlled by single pressure source. Such flexible, efficient, scalable, and low-cost design and formation methodology will accelerate the development and iteration of soft robots and also open up more opportunities for safe human-robot interactions, tailored wearable devices, and health care.