While preclinical stroke studies have shown that mesenchymal stem cells (MSCs) promote recovery, few randomized controlled trials (RCT) have assessed cell therapy in humans. In this RCT, we assessed the safety, feasibility, and efficacy of intravenous autologous bone marrow-derived MSCs in subacute stroke. ISIS-HERMES was a single-center, open-label RCT, with a 2-year follow-up. We enrolled patients aged 18-70 years less than 2 weeks following moderate-severe ischemic carotid stroke. Patients were randomized 2:1 to receive intravenous MSCs or not. Primary outcomes assessed feasibility and safety. Secondary outcomes assessed global and motor recovery. Passive wrist movement functional MRI (fMRI) activity in primary motor cortex (MI) was employed as a motor recovery biomarker. We compared "treated" and "control" groups using as-treated analyses. Of 31 enrolled patients, 16 patients received MSCs. Treatment feasibility was 80%, and there were 10 and 16 adverse events in treated patients, and 12 and 24 in controls at 6-month and 2-year follow-up, respectively. Using mixed modeling analyses, we observed no treatment effects on the Barthel Index, NIHSS, and modified-Rankin scores, but significant improvements in motor-NIHSS (p = 0.004), motor-Fugl-Meyer scores (p = 0.028), and task-related fMRI activity in MI-4a (p = 0.031) and MI-4p (p = 0.002). Intravenous autologous MSC treatment following stroke was safe and feasible. Motor performance and task-related MI activity results suggest that MSCs improve motor recovery through sensorimotor neuroplasticity. ClinicalTrials.gov Identifier NCT00875654.