The Coronavirus Disease 2019 (COVID-19) has roamed for almost two years now. Every country has applied its strategies in facing and handling this pandemic, including Indonesia. One strategy applied by the Indonesian government in handling this crisis is the enforcement of restrictions on community activities (PPKM) policy. This policy has been acknowledged by many countries’ leaders as an effective strategy in handling the COVID-19 pandemic without giving too much burden to the economic sector. However, despite the pros, there are also cons of the policy in society. Therefore, we are interested in conducting a sentiment analysis for the PPKM policy based on Twitter tweets data. We found that most of the tweets were dominated by the neutral sentiment (58.07%), followed by the positive sentiment (27.12%), and lastly by the negative sentiment (14.81%). Furthermore, we also try to build a deep learning model based on long short-term memory (LSTM) networks for the classification task of the collected tweets. We found the proposed deep learning model could reach 92.59% accuracy on the test set, which is pretty high for this sentiment analysis classification task. The built model then was deployed as a simple web-based application that can be accessed freely in the Heroku platform.