The work presented in this paper has addressed the issue of resource sharing in dynamic heterogeneous Multi Agent Systems as a search problem. When performing a random search, this might lead to traverse the whole network and increase the failure ratio. This paper has introduced heuristic directed search based on the usage of an approximate matching mechanism to overcome this problem. Our implementation of search algorithms differs from traditional algorithms by using semantically guided technique for resource search as well as a dynamically re-organisable network of agents. The experimental results have shown that using directed search techniques is better than random search in terms of number of hops to find the match. Furthermore, network re-organisation has improved the system performance by directing the search based on resources information, especially when high accuracy is required.