Dynamic　joins　and　leaves　of　the　node　and　heterogeneous　network　structure　in　P2P　system　make　it　a　difficult　problem　to　discover　shared　resource.　A　resource　discovery　model　based　on　the　multi-agent　technology　is　presented.　Agents　maintain　local　resource　information　and　mobile　agents　migrate　on　network　to　discover　shared　resource.　Multi-agents　cooperate　with　each　other　by　the　mean　of　pheromone　and　local　heuristic　information　in　Ant　System.　System　model　and　decision　algorithm　is　described.　Resource　can　be　quickly　acquired　and　network　traffic　can　be　reduced　greatly,　scalability　and　robustness　can　also　be　improved　based　on　this　model　in　P2P　system.　©　2004　IEEE.