In　view　of　characteristics　of　flexible　job　shop　scheduling,　a　distributed　agent-based　flexible　scheduling　system　is　proposed.　The　system　contains　job　agents,　unit　agents　and　a　global　agent.　Each　relevant　agent　calculates　an　agent　priority　for　competing　jobs.　The　scheduling　system　determines　each　agent＇s　weight　based　on　system　load　and　job　due　date,　and　obtains　final　priority　as　a　weighted　average　of　agent　priorities.　The　system　is　implemented　with　Arena　and　simulated　under　various　flexible　job　shop　settings　with　weighted　tardiness　objective.　The　results　show　that　the　proposed　agent-based　scheduling　method　significantly　outperforms　traditional　dispatching　rules.