GitHub’s Octoverse 2026 report points to a sharp rise in AI agent activity: by September, agents initiated more than 40% of pull requests, up from 17% in March, totaling 110 million that month—about six times March’s volume.

That surge reflects broader adoption of agentic tools: 55% of GitHub Copilot monthly active users used agentic features, up from 35% a year earlier, while use of Copilot’s cloud agent grew 230% year over year.

Agent-started pull requests were more likely to touch multiple files than other pull requests, at 87% versus 79%, and 90% merged within two weeks, compared with about 86% of the comparison group.

Meanwhile, paid Copilot users selected from 47 distinct models in September, up from 29 in May, with no single model accounting for more than 25% of those requests; the figures exclude automatic model selection and requests made with users’ own API keys.

GitHub cautions that its agent-use indicators are a minimum because some activity cannot be detected, and the report does not name the vendors or models behind agent-started pull requests; the figures come from GitHub telemetry that the article could not independently verify.

The report also captures GitHub’s overall scale: in the 12 months ending September 30, developers pushed code 17.9 billion times, made 34.3 billion commits, merged 1.3 billion pull requests and ran 45.4 billion GitHub Actions jobs, while GitHub says it serves more than 225 million developers.