Shortages　of　traditional　intelligent　agent　searching　are　slow　speed　and　having　robustness　problems,　because　of　the　large　amount　of　information　to　exchange.　Therefore,　it　cannot　meet　the　requirements　of　society　now.　In　order　to　improve　the　search　efficiency　and　fault　tolerance,　this　paper　studies　the　small　world　network　model　of　Kleinberg,　and　finds　this　kind　of　network　structure　can　be　contained　in　a　bounded　path　length.　Kleinberg＇s　model　is　the　foundation　of　the　proposed　solution.　Forming　a　navigable　small　world　network　and　providing　an　external　view　of　a　consistent　entity　as　a　unified　service　by　AMS　and　DF.　The　result　indicate　that　the　proposed　navigable　network　can　reduce　the　scale　of　information　exchange　from　n(2)　to　2.5(logn)(2),　and　keep　the　search　failure　rate　below　10%.