Automaton-based pattern matching methods have been widely used in security services for traffic inspection and filtering. However, a large scale of patterns may be updated frequently in a multi-tenant cloud, which poses new challenges to avoid attacks while updating new patterns. This paper presents an agile approach named Eagle for "on-the-fly" updating automaton in cloud security services. The approach provides three algorithms on AC and SBOM, adding, deleting and updating operation, to update state and links of automaton in high-speed online cloud traffic. Theoretical analysis shows that Eagle lowers the computational complexity of updating patterns from O (n2) to O (n). The effectiveness of this agile approach is verified when applied to a real cloud gateway. It turns out that 68% - 89% of the time can be saved and the throughput of cloud traffic filtering proves no reduction during and after the pattern update.