Six months, 17% to 40%
More than 40% of pull requests opened on GitHub in September 2026 started with an agent’s first draft, up from 17% in March, according to this year’s Octoverse report, published on 28 October. In absolute terms that is 110 million agent-started pull requests in a single month, roughly six times the March volume.
The figures are GitHub’s own, drawn from activity on its platform between 1 October 2025 and 30 September 2026, and they are a measurement of where code changes begin rather than of who wrote them. A pull request that an agent drafts is still reviewed, amended and merged by people — or not merged at all.
The drafts are bigger, and they land
Two of the report’s secondary numbers say more about the shift than the headline share does. About 87% of agent-started pull requests touched more than one file, against 79% of pull requests with no detected agent involvement: the drafts are not single-line fixes.
And they are being accepted. Pull requests with agent involvement reached a 90% merge rate within two weeks in September 2026, up from 83% in February. The comparison group stayed at about 86% across both months. A year ago agent-drafted changes merged more slowly than human-authored ones; now they merge faster.
The totals moved with them
The platform-wide counters rose steeply. GitHub recorded 34.3 billion commits in the twelve months, against 12.9 billion the year before and 10.0 billion in 2024. Pushes reached 17.9 billion, from 6.7 billion. Merged pull requests came to 1.3 billion, about 3.5 million a day, from 538 million. GitHub Actions ran 45.4 billion jobs, roughly 1,440 a second.
One correlation in the report is worth separating from its claim: among repositories created this year, those containing an agent configuration file averaged eight times as many commits as those without one. That measures which repositories are busy, not what caused the activity.
GitHub now counts more than 225 million developers, with 38.3 million in the United States, 32.0 million in India, 12.4 million in China and 10.2 million in Brazil. TypeScript ranks first by monthly contributors, ahead of Python and JavaScript. GitHub puts AI involvement in merged pull requests at 44.1% for TypeScript, 41.1% for Python, 40.4% for Rust and 21.2% for Java.
Developers are no longer picking one model
Inside Copilot, 55% of monthly active users now use agentic features, up from 35% a year earlier, and cloud agent usage grew 230% year on year. Among users who chose a model themselves in September, 61% used more than one, up from 50% four months earlier. Paying users touched 47 distinct models in the month, against 29 in May, and no single model accounted for more than 25% of requests.
What to watch
The agent-started share is the figure to follow, because the comparison it invites is now straightforward: if the merge-rate gap holds as volume grows, review capacity rather than drafting becomes the constraint on how fast software ships. GitHub publishes Octoverse annually, which means the next directly comparable number is a year away.
