In this letter, we will discuss the design, modelling, and estimation of load based on the shape of a newly developed parallel continuum robot called TriRod. The proposed kinematic architecture consists of three flexible rods attached at their lower extremity to three rotating actuators. The robot is able to move its end-effector in space and its three limbs are designed so that their shapes can easily be visible by cameras. Two shape-based estimation strategies have been developed and evaluated experimentally using the TriRod robot. Both estimators are based on virtual visual servoing of the robot rod shapes (deformation) to estimate either the joint angular values or the external forces. The experimental validations show that both estimators give accurate predictions of angles and forces, with a mean average error of 3.8% for joint value estimation over a range of 75 ± and a mean average error of 10.5% for the estimation of loads up to 633mN .