More than 40% of GitHub pull requests in September began with an AI agent's first draft, as the software platform logged record activity across the year ended Sept. 30.
GitHub put September's count of such requests at 110 million, roughly six times the number recorded in March. The share rose from 17% in March to above 40% in September, according to the company's 2026 Octoverse report. GitHub said activity picked up around June as models improved, agent tools and workflows expanded, and cloud execution gained ground. It did not assign the increase to any one factor.
During the 12 months from Oct. 1, 2025, through Sept. 30, 2026, GitHub logged 17.9 billion pushes, 34.3 billion commits and 1.3 billion merged pull requests. Merged pull requests averaged 3.5 million a day. The push count topped the combined 2024 and 2025 total after GitHub recalculated both years using matching definitions and a 12-month window. The report's activity totals generally include public and private repositories. One push may carry several commits, so the two figures track different kinds of activity.
GitHub's engineering team separately reported 7.38 billion commits in September, more than five times the September 2025 count. The team also said GitHub is rebuilding its Git infrastructure to support developers and AI agents working concurrently.
The report also grouped repositories created during the year by whether they included an agent configuration file. The configured group averaged eight times the commit volume of the group without the file, though that comparison alone does not show the configuration drove the gap.
At GitHub Universe, Microsoft Executive Vice President of CoreAI Jay Parikh framed the goal around finished work: “Completed work, not more code, is the output that matters.” The Octoverse figures count platform activity and do not assess the quality or usefulness of the software those actions produce.
