Abstract: We propose an appearance-based facial recognition system based on multi-agents. By this method a person can be identified and verified using his face. Each RETSINA agent performs the facial recognition with its own small training sets and its stops other agents when it finds a match by communicating through ACL. Experimental results suggest that the proposed multi-agent based Facial recognition method provides a better recognition rate and achieves less time complexity. The worst case analysis shows proposed multi-agent based facial recognition is 80% better than ordinary facial recognition.