Taking vehicle queue length in urban intersection, traffic flow and signal timing at the present situation as input, the intersection signal timing scheme as output, the minimum network delay as the goal, it constructs a multi - agent traffic signal control system including the execution layer, control layer and decision-making layer. When the queue length in intersection is higher than the setting threshold, the intersection agent is activated. By sending coordination request to the adjacent intersections, based on the NASH equilibrium theory, it looks for the signal timing scheme that is satisfied with the minimum delay time to coordinate intersection agent. If NASH equilibrium has no solution, the network upgrades for management agent coordination and conducts area agent timing coordination according to the historic timing scheme stored in rule base. The simulation analysis shows that TSCA (Traffic Signal Control Agent) system can effectively reduce the overall delay time of the system and the control effect is better.