Presently,　there　are　still　many　problems　on　acquiring　patent　information　from　the　Website　,we　can＇t　acquire　a　large　number　of　information　efficiently　and　accurately.　In　this　paper,　use　the　two-tiered　scheduling　Multi-Agent　to　design　the　patent　information　acquisition　system.　The　process　of　acquisition　patent　information　from　United　States　Patent　and　Trademark　Office　(USPTO)　web　database　which　includes　patent　searching,　interface　analysis,　patent　download　and　information　extraction　has　been　studied.　The　methods　such　as　Deep　Web　technology,　Multi-Agents　technology　and　information　extraction　technology　based　on　XML　have　been　used　in　patent　acquisition.　The　purpose　of　the　paper　is　acquiring　patent　information　accurately,　rapidly　and　automatically　from　USPTO　database.　©2010　IEEE.