Labor shortage and high labor costs are critical issues in agriculture that have been addressed through the design of automatic harvesting robots. In the case of tomatoes, the second most widely consumed crop worldwide, several proposals have been presented, with some reaching a commercial Technology Readiness Level. However, their selling prices required to implement these solutions hinder scalability, particularly in small and medium-sized agricultural segments. This work presents a proposed development of a low-cost autonomous harvesting robot for greenhouse tomatoes. Conceptually, a gantry-based mechatronic system, equipped with a YOLOv8-based computer vision system and a flexible end-effector is designed to detect, locate, and pick the harvestable tomatoes. Preliminary results, obtained at a laboratory research level, validate the functionalities of the prototype. In particular, the computer vision system detects harvestable tomatoes with a precision of 97%, while the tomatoes are picked in an average time of less than 5 s.