To address the challenges in oil and gas pipeline corrosion management, such as the fragmentation of digital assets and the limitations of manual risk assessment, a method for identifying corrosion risk points based on agent-knowledge graph synergy is proposed. Through spatial and temporal alignment, the scattered digital assets in the corrosion management process are systematically structured. In accordance with industry standards, a multimodal corrosion knowledge base ontology is designed to extract and integrate heterogeneous data. A corrosion risk point identification agent is constructed using a large language model (LLM), and a corrosion mechanism quantification toolset is developed by importing corrosion mechanisms and standards. The corrosion risk assessment logic and corrosion risk evaluation are determined by key semantic information, operational conditions, historical corrosion case databases. Full utilization of dispersed and heterogeneous data is enabled by the proposed method, facilitating efficient corrosion risk identification. The effectiveness of the proposed method has been validated at a natural gas gathering and transmission station of Chongqing gas field.