Traditional software process models are mostly static, mechanical, and passive. Traditional approach requires modeler to determine all the possible conditions the software process will encounter and to define explicitly the solutions into a process model. It lacks the ability to allow further deliberations when the modeled environment changes. This paper presents an Agent-based self-adaptive software process model. In this approach, software process is modeled as peers: process Agents. These software process Agents can adapt themselves to the software process environment and act with initiative and autonomy. When the process environment changes, the process agents can dynamically change their behavior to ensure that the development goal can still be achieved.