This study presents effects of adding Circle of Security-Parenting (COS-P) to an already established comprehensive therapeutic model for early parent-child intervention in three Swedish infant mental health (IMH) clinics. Parents' internal representations and quality of parent-infant interaction were studied in a clinical sample comprised of 52 parent-infant dyads randomly allocated to two comparable groups. One group consisted of 28 dyads receiving treatment as usual (TAU) supplemented with COS-P in a small group format, and another group of 24 dyads receiving TAU only. Assessments were made at baseline (T1), 6 months after inclusion (T2) and 12 months after inclusion (T3). Changes over time were explored in 42 dyads. In the COS-P group, the proportion of balanced representations, as assessed with Working Model of the Child Interview (WMCI), significantly increased between T1 and T3. Further, the proportion of emotionally available interactions, as assessed with Emotional Availability scales (EA), significantly increased over time in the COS-P group. Improvements in the TAU-group were close to significant. Limitations of the study are mainly related to the small sample size. Strength is the real world character of the study, where COS-P was implemented in a clinical context not otherwise adapted to research. We conclude by discussing the value of supplementing TAU with COS-P in IMH treatment.