Persons post-stroke experience excessive muscle co-contraction, and consequently the arm functions are compromised during the activities of daily living. Therefore, identifying instrumental outcome measures able to detect the motor strategy adopted after a stroke is a primary clinical goal. Accordingly, this study aims at verifying whether the surface electromyography (sEMG)-based co-contraction index (CCI) could be a new clinically feasible approach for assessing and monitoring patients’ motor performance. Thirty-four persons post-stroke underwent clinical assessment and upper extremity kinematic analysis, including sEMG recordings. The participants were randomized into two treatment groups (robot and usual care groups). Ten healthy subjects provided a normative reference (NR). Frost’s CCI was used to quantify the muscle co-contraction of three different agonist/antagonist muscle pairs during an object-placing task. Persons post-stroke showed excessive muscle co-contraction (mean (95% CI): anterior/posterior deltoid CCI: 0.38 (0.34–0.41) p = 0.03; triceps/biceps CCI: 0.46 (0.41–0.50) p = 0.01) compared to NR (anterior/posterior deltoid CCI: 0.29 (0.21–0.36); triceps/biceps CCI: 0.34 (0.30–0.39)). After robot therapy, persons post-stroke exhibited a greater improvement (i.e., reduced CCI) in proximal motor control (anterior/posterior deltoid change score of CCI: −0.02 (−0.07–0.02) p = 0.05) compared to usual care therapy (0.04 (0.00–0.09)). Finally, the findings of the present study indicate that the sEMG-based CCI could be a valuable tool in clinical practice.