Service networks made up of nodes and links inevitably suffer from performance degradation due to the negative effect of natural disasters and intentional attacks. Resilience is defined as the capability of recovering from disruptive events. Resilience analysis is of vital importance to evaluate network performance during the whole operation process. Given the dynamic characteristic of service networks, it is difficult to reflect the actual performance through static models. Moreover, resilience assessment should proceed with multiple metrics because network resilience is a multifaceted capability. To this end, an agent-based framework for resilience analysis is developed in this paper. Nodes are regarded as agents with independent decision-making to better respond to disruptions. The multi-agent negotiation mechanism is introduced to satisfy service requirements using deep Q-learning. In addition, network resilience is comprehensively assessed in terms of reliability, supportability and maintainability. A case study of Iridium system is conducted to illustrate the applicability of the agent-based framework. The results show that the developed framework can select the optimal route for task assignment and quantify resilience in the dynamic environment.