BACKGROUND: Recovery features strongly in Australian mental health policy; however, evidence is limited for the efficacy of recovery-oriented practice at the service level. This paper describes the Principles Unite Local Services Assisting Recovery (PULSAR) Specialist Care trial protocol for a recovery-oriented practice training intervention delivered to specialist mental health services staff. The primary aim is to evaluate whether adult consumers accessing services where staff have received the intervention report superior recovery outcomes compared to adult consumers accessing services where staff have not yet received the intervention. A qualitative sub-study aims to examine staff and consumer views on implementing recovery-oriented practice. A process evaluation sub-study aims to articulate important explanatory variables affecting the interventions rollout and outcomes. METHODS: The mixed methods design incorporates a two-step stepped-wedge cluster randomized controlled trial (cRCT) examining cross-sectional data from three phases, and nested qualitative and process evaluation sub-studies. Participating specialist mental health care services in Melbourne, Victoria are divided into 14 clusters with half randomly allocated to receive the staff training in year one and half in year two. Research participants are consumers aged 18-75 years who attended the cluster within a previous three-month period either at baseline, 12 (step 1) or 24 months (step 2). In the two nested sub-studies, participation extends to cluster staff. The primary outcome is the Questionnaire about the Process of Recovery collected from 756 consumers (252 each at baseline, step 1, step 2). Secondary and other outcomes measuring well-being, service satisfaction and health economic impact are collected from a subset of 252 consumers (63 at baseline; 126 at step 1; 63 at step 2) via interviews. Interview-based longitudinal data are also collected 12 months apart from 88 consumers with a psychotic disorder diagnosis (44 at baseline, step 1; 44 at step 1, step 2). cRCT data will be analyzed using multilevel mixed-effects modelling to account for clustering and some repeated measures, supplemented by thematic analysis of qualitative interview data. The process evaluation will draw on qualitative, quantitative and documentary data. DISCUSSION: Findings will provide an evidence-base for the continued transformation of Australian mental health service frameworks toward recovery. TRIAL REGISTRATION: Australian and New Zealand Clinical Trial Registry: ACTRN12614000957695 . Date registered: 8 September 2014.