This thesis proposes a message security solution on multi-agent systems. A general security analysis based on properties of software agents is presented along with an overview of security measures applicable to multi-agent systems. A security design and implementation has been developed to protect communication among agents. And this implementation scheme has been applied to Seagent, a semantic web enabled multi-agent framework. Hence, a set of agent security mechanisms have been adapted for Seagent and have been implemented for message confidentiality, integrity, authentication and nonrepudiation. Then these mechanisms have been tested for communication performance on Seagent.