Mild frailty or pre-frailty is common and yet is potentially reversible. Preventing progression to worsening frailty may benefit individuals and lower health/social care costs. However, we know little about effective approaches to preventing frailty progression. (1) To develop an evidence- and theory-based home-based health promotion intervention for older people with mild frailty. (2) To assess feasibility, costs and acceptability of (i) the intervention and (ii) a full-scale clinical effectiveness and cost-effectiveness randomised controlled trial (RCT). Evidence reviews, qualitative studies, intervention development and a feasibility RCT with process evaluation. Two systematic reviews (including systematic searches of 14 databases and registries, 1990-2016 and 1980-2014), a state-of-the-art review (from inception to 2015) and policy review identified effective components for our intervention. We collected data on health priorities and potential intervention components from semistructured interviews and focus groups with older people (aged 65-94 years) ( = 44), carers ( = 12) and health/social care professionals ( = 27). These data, and our evidence reviews, fed into development of the HomeHealth intervention in collaboration with older people and multidisciplinary stakeholders. HomeHealth comprised 3-6 sessions with a support worker trained in behaviour change techniques, communication skills, exercise, nutrition and mood. Participants addressed self-directed independence and well-being goals, supported through education, skills training, enabling individuals to overcome barriers, providing feedback, maximising motivation and promoting habit formation. Single-blind RCT, individually randomised to HomeHealth or treatment as usual (TAU). Community settings in London and Hertfordshire, UK. A total of 51 community-dwelling adults aged ≥ 65 years with mild frailty. Feasibility - recruitment, retention, acceptability and intervention costs. Clinical and health economic outcome data at 6 months included functioning, frailty status, well-being, psychological distress, quality of life, capability and NHS and societal service utilisation/costs. We successfully recruited to target, with good 6-month retention (94%). Trial procedures were acceptable with minimal missing data. Individual randomisation was feasible. The intervention was acceptable, with good fidelity and modest delivery costs (£307 per patient). A total of 96% of participants identified at least one goal, which were mostly exercise related (73%). We found significantly better functioning (Barthel Index +1.68; = 0.004), better grip strength (+6.48 kg; = 0.02), reduced psychological distress (12-item General Health Questionnaire -3.92; = 0.01) and increased capability-adjusted life-years [+0.017; 95% confidence interval (CI) 0.001 to 0.031] at 6 months in the intervention arm than the TAU arm, with no differences in other outcomes. NHS and carer support costs were variable but, overall, were lower in the intervention arm than the TAU arm. The main limitation was difficulty maintaining outcome assessor blinding. Evidence is lacking to inform frailty prevention service design, with no large-scale trials of multidomain interventions. From stakeholder/public perspectives, new frailty prevention services should be personalised and encompass multiple domains, particularly socialising and mobility, and can be delivered by trained non-specialists. Our multicomponent health promotion intervention was acceptable and delivered at modest cost. Our small study shows promise for improving clinical outcomes, including functioning and independence. A full-scale individually RCT is feasible. A large, definitive RCT of the HomeHealth service is warranted. This study is registered as PROSPERO CRD42014010370 and Current Controlled Trials ISRCTN11986672. This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in ; Vol. 21, No. 73. See the NIHR Journals Library website for further project information.