Additive manufacturing (AM) is an automated process for fabricating complex structures, especially for continuous fiber-reinforced polymer (CFRP) composites with exceptional mechanical performance and reduced weight. The extrusion-based CFRP-AM process has been widely adopted but faces challenges with insufficient impregnation/adhesion and lack of geometrical accuracy. Existing research efforts exploit various mechatronics approaches and printing parameter adjustment techniques to enhance CFRP quality. Nevertheless, most of them are static or slow-varying; thus, they cannot dynamically adjust to local geometrical features and faults, which are particularly representative of the CFRP-AM process in high-curvature regions. To enable dynamic and rapid adjustment, this study exploits contact forces as major real-time feedback signals to dynamically control the feed velocity, layer height, and fiber extrusion rate. To create a continuous feedback control environment for dynamic adjustment, the study also proposes a novel helical CFRP-AM printing trajectory generation method for closed shell structures, such that the helical angles indirectly adjust the layer height of the parts. A material extrusion CFRP-AM platform combining a co-extrusion nozzle, an industrial robot, a six-degree-of-freedom load cell, and a real-time closed-loop control system is built to verify the proposed control framework. Three representative case studies are introduced to show the enhancements of the closed-loop CFRP-AM process, particularly in layer height adjustment, high-speed printing, and large curvature regions.