The prediction of tool life is of great significance to ensure the quality of parts and control the cost of machining. However, the tool wear process is complex and changeable, and it is difficult to accurately predict the residual life of the cutting tools affected by machining conditions. To solve the above problems, this paper presents a dynamic prediction method of tool life based on online learning. Using long-short term memory as base model and integrating the online learning module, the final model can automatically update the parameters during the machining process, and the accurate prediction of tool life under variable working conditions can be realized. The milling experiment was carried out, and the experimental results show that the dynamic prediction method of tool life can effectively improve the precision of tool life prediction.