Abstract Background Two thirds of survivors will achieve independent ambulation after a stroke, but less than half will recover upper limb function. There is strong evidence to support intensive repetitive task-oriented training for recovery after stroke. The number of repetitions needed is suggested to be in the order of hundreds, but this is not currently being achieved in clinical practice. In an effort to bridge this evidence-practice gap, we have developed a behaviour change intervention that aims to increase provision of upper limb repetitive task-oriented training in stroke rehabilitation. This paper aims to describe the systematic processes that took place in collaboratively developing the behaviour change intervention. Methods The methods used in this study were not defined a priori but were guided by the Behaviour Change Wheel. The process was collaborative and iterative with four stages of development emerging (i) establishing an intervention development group; (ii) structured discussions to understand the problem, prioritise target behaviours and analyse target behaviours; (iii) collaborative design of theoretically underpinned intervention components and (iv) piloting and refining of intervention components. Results The intervention development group consisted of the research team and stroke therapy team at a local stroke rehabilitation unit. The group prioritised four target behaviours at the therapist level: (i) identifying suitable patients for exercises, (ii) provision of exercises, (iii) communicating exercises to family/visitors and (iv) monitoring and reviewing exercises. It also provides a method for self-monitoring performance in order to measure fidelity. The developed intervention, PRACTISE (Promoting Recovery of the Arm: Clinical Tools for Intensive Stroke Exercise), consists of team meetings and the PRACTISE Toolkit (screening tool and upper limb exercise plan, PRACTISE exercise pack and an audit tool). Conclusions This paper provides an example of how the Behaviour Change Wheel may be applied in the collaborative development of a behaviour change intervention for health professionals. The process involved was resource-intensive, and the iterative process was difficult to capture. The use of a published behaviour change framework and taxonomy will assist replication in future research and clinical use. The feasibility and acceptability of PRACTISE is currently being explored in two other stroke rehabilitation units.