GHARuns: A Large Dataset of GitHub Actions Workflow Execution Metadata

GitHub Actions is the integrated CI/CD workflow automation service of GitHub, the most popular collaborative software development platform. While the evolution of workflow configuration files has been actively studied, there are considerably less large-scale empirical studies involving execution and runtime behaviour of those workflows. This is due to the data retention policy imposed by GitHub and the resulting lack of longitudinal data involving large numbers of repositories. We propose GHARuns, an evolving dataset and associated extraction tool to enable such large-scale studies. Version 2026-07-08 of the dataset contains the metadata of 71.7M+ workflow runs for 197K+ GitHub Actions workflows across 45K+ active public GitHub repositories over a period of 146 days, well exceeding GitHub’s retention period. It also contains the metadata for for 122.7M+ job runs and 1362.3M+ step runs for a subset of 29.4M+ workflow runs.