Collaborative software development through GitHub repositories frequently relies on bot accounts that automate many repetitive and error-prone tasks. Several bot identification tools and techniques have been proposed in the past, but they tend to rely on a substantial amount of historical data, or they limit themselves to a reduced subset of activity types. To overcome these limitations, we developed RABBIT, an open-source command-line tool that queries the GitHub Events API to retrieve the recent events of a given GitHub account to predict whether the account is a human or a bot. The prediction is based on an XGBoost classification model that relies on six features related to account activities, and that is obtained through grid-search 10-fold cross-validation. Based on a newly created ground-truth dataset of GitHub accounts containing 644 bots and 691 humans, we trained the model on 60% of the data and achieved a very good performance, with an AUC of 0.97, weighted F1 score of 0.94, precision of 0.94 and recall of 0.94. After integrating this model into RABBIT, and testing the tool’s performance on the remaining 40% of unseen data achieved a performance of 0.93 for AUC, weighted F1 score, precision and recall each. Taking into account the imposed hourly GitHub API limit, RABBIT can classify thousands of accounts per hour using at most 3 queries per account.