Actively engaging communities in effective partnerships for the design and implementation of HIV prevention research is vital to the successful conduct of ethically robust, locally-appropriate clinical trials in developing countries. This is especially true in vulnerable at-risk sub-populations, where definitions of "community", "participation" and "representation" can be difficult to apply. This study was conducted to investigate the feasibility of a participatory model of community liaison among an occupational cohort of women at high-risk of HIV and sexually-transmitted infections in Mwanza City, northwest Tanzania in preparation for a Phase III vaginal microbicide trial. This approach was rooted in participatory action-orientated research and used tools adapted from participatory learning and action techniques. During the feasibility study, a mobile community-based sexual and reproductive health service for women working as informal food vendors or in traditional and modern bars, restaurants, hotels and guesthouses was established in 10 city wards. Participatory mapping was carried out by project fieldworkers and wards divided into 78 geographical clusters of facilities in consultation with community members and study participants. Representatives at cluster and ward level were elected in a process facilitated by the site Community Liaison Officer and a site-level Community Advisory Committee established. A logical framework was used to guide the implementation, monitoring and evaluation of the community liaison system (CLS) within the broader feasibility study. The CLS was essential to the successful conduct of the feasibility study and has now been consolidated and expanded as part of the on-going MDP301 Phase III microbicide trial in Mwanza. The participatory model presented in this paper is likely to be generalisable to other vulnerable, stigmatised, at-risk study populations in resource-limited settings.