Intelligence plays important role in network management (NM), but traditional centralized NM framework has limited feasibility of implementation for it with core being management information base (MIB). AGIMA is one distributed computing centered NM prototype system based on intelligent agent (IA) group. The modal logic, which was applied by early scholars when analyzing multi-agent knowledge system, has many drawbacks, while this paper exploits high-level Petri net graph for modeling AGIMA. After giving net graphs of agent and agent group, authors present their mappings to a synchronize and concurrent behavior and show their capacities of analysis and verification of properties within distributed knowledge base. Easy migration to object and layered model also brings net graph auxiliary functions to system implementation.