The locations of WiFi access points (APs) are important for WiFi positioning, especially when a propagation model is used. The parameters for the propagation model, such as the pathloss exponent and noise variance, usually are not available when localising APs in a new environment. A crowdsourcing-based prototype system is introduced that automatically generates WiFi databases using the uploaded data during normal usage of smartphones. In this system, the adjustment algorithm is originally used for the estimation of AP localisation and propagation parameters. Preliminary experiments show that the average AP localisation error of the prototype system is about 4.0 m in a typical indoor environment with considerably reduced time and labour costs compared with traditional methods.