Introduction Cerebral palsy (CP) is the most common childhood physical disability, with 80%estimated to be in low-middle-income countries. This study aims to (1) determine the accuracy of General Movements (GMs)/Hammersmith Infant Neurological Examination (HINE) for detecting CP at 18 months corrected age (CA); (2) determine the effectiveness of a community-based parent-delivered early intervention for infants at high risk of CP in West Bengal, India (Learning through Everyday Activities with Parents for infants with CP; LEAP-CP). Methods This study comprises two substudies: (1) a study of the predictive validity of the GMs and HINE for detecting CP; (2) randomised, double-blinded controlled trial of a novel intervention delivered through peer trainers (Community Disability Workers, CDW) compared with health advice (15 fortnightly visits). 142 infants at high risk of CP ('absent fidgety' GMs; 'high risk score' on HINE) aged 12-40 weeks CA will be recruited to the intervention substudy, with infants randomised based on a computer-generated sequence. Researchers will be masked to group allocation, and caregivers and CDWs naïve to intervention status. Visits will include therapeutic modules (goal-directed active motor/cognitive strategies and LEAP-CP games) and parent education. Health advice is based on the Integrated Management of Childhood Illness, WHO. Infants will be evaluated at baseline, post intervention and 18 months CA. The primary hypothesis is that infants receiving LEAP-CP will have greater scaled scores on the Pediatric Evaluation of Disability Inventory-Computer Adaptive Test (mobility domain) at 18 months compared with health advice. Secondary outcomes include infant functional motor, cognitive, visual and communication development; infant growth; maternal mental health. Ethics and dissemination This study is approved through appropriate Australian and Indian ethics committees (see in text) with families providing written informed consent. Findings from this trial will be disseminated through peer-reviewed journal publications and conference presentations. Trial registration number 12616000653460p; Pre-results.