Self-reported subjective cognitive difficulties (subjective deficits) and rumination are central residual cognitive symptoms following major depressive disorder (MDD). These are risk factors for more a severe course of illness, and despite the considerable relapse risk of MDD, few interventions target the remitted phase, a high-risk period for developing new episodes. Online distribution of interventions could help close this gap. Computerized working memory training (CWMT) shows promising results, but findings are inconclusive regarding which symptoms improve following this intervention, and its long-term effects. This study reports results from a longitudinal open-label two-year follow-up pilot-study of self-reported cognitive residual symptoms following 25 sessions (40 min), five times a week of a digitally delivered CWMT intervention. Ten of 29 patients remitted from MDD completed two-year follow-up assessment. Significant large improvements in self-reported cognitive functioning on the behavior rating inventory of executive function-adult version appeared after two-years (d = 0.98), but no significant improvements were found in rumination (d < 0.308) measured by the ruminative responses scale. The former showed moderate non-significant associations to improvement in CWMT both post-intervention (r = 0.575) and at two-year follow-up (r = 0.308). Strengths in the study included a comprehensive intervention and long follow-up time. Limitations were small sample and no control group. No significant differences between completers and drop-outs were found, however, attrition effects cannot be ruled out and demand characteristics could influence findings. Results suggested lasting improvements in self-reported cognitive functioning following online CWMT. Controlled studies with larger samples should replicate these promising preliminary findings.