Robot-assisted rehabilitation of hand function is becoming an established approach to complement conventional therapy after stroke, particularly in view of its possible unsupervised use to promote an increase in therapy dose. Given their intensive therapy regime, robots may promote a temporary increase in hand muscle tone and spasticity, which may cause pain and negatively affect recovery. To integrate hand muscle tone monitoring into an assessment-driven robot-assisted therapy concept, an online assessment of muscle tone is proposed and incorporated into an exercise. The exercise was preliminarily tested in a pilot study with five chronic stroke survivors (non-spastic at rest) and five healthy participants to identify the range of potential physiological muscle tone change that can happen also in a non-spastic population during a single exercise session. In both groups, the muscle tone level during hand opening was higher in fast 20 mm ramp-and-hold perturbations (150 ms) compared to slow (250 ms) perturbations, and corresponded to a force change of approximately 4-5 N. Despite not being statistically significantly different, in the stroke group the force change (and the speed dependency) increased with exercise time. This information could serve as a basis to develop strategies to continuously adapt the difficulty and activity level required in robot-assisted rehabilitation and to monitor or even control the muscle tone evolution over time.