The client-therapist relationship has long been recognized as an important element in psychotherapy, and research has demonstrated its robust association with positive outcomes. This study examined the feasibility and preliminary efficacy of training therapists in strategies for improving therapeutic relationships with clients. The strategies were compiled from the empirical literature, drawing on the work of Hilsenroth and Cromer (2007), Castonguay (1996), and Safran and Muran (2000). The study employed a manipulated training design that has the benefit of addressing naturalistic effectiveness questions, while adhering to the rigorous scientific standards of controlled efficacy research (Hayes, 2002). Participants were 57 therapists working at five community mental health clinics who were randomly assigned to the brief alliance training workshop (in which they participated prior to starting treatment with a new client) or to a delayed-training control condition. Outcomes assessed included therapists' self-reported use of alliance strategies in session 1, therapist-rated alliance quality after session 1, and early client engagement. Engagement was operationalized in several ways: number of sessions attended in the first four weeks, planned session frequency (e.g., weekly, monthly), attendance rate (i.e., percent of scheduled sessions attended), and treatment status at the end of four weeks (e.g., therapist and client had next session scheduled, client had terminated unilaterally). Counter to hypotheses, one-way ANOVAs and chi-square analyses revealed no statistically significant differences between the training and the delayed-training conditions on the primary outcomes. However, effect size estimates suggested that clinicians in the training condition reported better alliances with their clients than clinicians who had yet to receive the training (d = 0.40, 95% CI [-0.13, 0.93], small to medium effect). Furthermore, therapists' use of alliance strategies taught in the workshop was significantly correlated with alliance quality. In addition to the preliminary efficacy findings, the study generated important information about the feasibility of conducting psychotherapy research in naturalistic settings, as well as recommendations for future studies. The manipulated training design holds promise for collaborations between researchers and clinicians seeking to bridge science and practice.