Don Hoi Lord, a coastal wetland in the upper gulf of Thailand, is famous for the razor clam (Solen regularis), which is a source of food and income for local fishermen. However, the razor clam population is significantly decreasing so the aim of this study was to analyze the razor clam fishery and develop a management plan engaging the different stakeholders in collective discussions. A participatory modelling approach was used, including (a) agent-based modelling (ABM) and (b) role-playing games (RPG). First, an individual-state distribution model of the razor clam population was designed and calibrated to synthesise knowledge. Two RPG sessions supported by this biological model were organised to initiate collective learning and promote discussion among stakeholders. After playing with the baseline scenario showing the actual harvesting situation, discussions led to the definition of alternative scenarios for clam management such as establishment and rotation of zones closed for harvesting, and quota systems. These scenarios developed during the RPG sessions were more deeply investigated with an agent-based model that included fishermen agents and the individual-state distribution model of the razor clam population. Simulations of different scenarios (reserve, quota and combination of reserve and quota) were produced to enable better informed discussion between different stakeholders. Stakeholders have acknowledged that management through rotating reserves is not efficient if the duration of reserves is too short. They have realised that the efficiency of reserves is hard to assess without better data on clam recruitment. They have also explored another management option which is the quota system. In the final stage, the local participatory process was endorsed by the incumbent regional policy-making actor. Unfortunately, because of rapid administrative rotation the project abruptly lost support. This unexpected event emphasises the need for continuous monitoring o