Octoverse: Commits, pushes, reach new highs in 2026 as developers build with agents

What it claims

GitHub's 2026 Octoverse report is first-party telemetry covering the 12 months to about September 2026. It goes well beyond push counts. Commits across public and private repos reached 34.3 billion, more than three times the figure two years earlier. Merged pull requests more than tripled to 1.3 billion, about 3.5 million a day. GitHub dates a sharp surge in activity to around June 2026. It links the surge to agents becoming persistent across days, orchestrating other agents, and running in the cloud as well as on local machines. GitHub now counts more than 225 million developers.

Evidence

Agent-drafted pull requests. About 110 million pull requests started with an agent's first draft in September 2026, about 6x March's volume. Of agent-started PRs, 87% change multiple files, against 79% of PRs without detected agent involvement. Merge rates within two weeks shifted over the year:

• February: 83% for agent PRs, about 86% for the comparison group.
• September: 90% for agent PRs, while the comparison group held at about 86%.

Agents and repo activity. Among repos created in 2026, those with an agent configuration file averaged eight times as many commits as those without one. From June onward, about 55% of all commits came from brand-new repositories.

Copilot usage. 55% of Copilot monthly active users use agentic features, up from 35% a year earlier. Copilot cloud agent, which takes an assigned task and works on it alone, grew 230% year over year. Agent chat grew 85%.

Model selection. In September 2026, 61% of Copilot users who picked a model themselves used more than one, up from 50% four months earlier. Paid-plan users touched 29 distinct models in May and 47 in September. No single model carried more than 25% of requests. Developer-selected requests split 38% general purpose, 37% deep reasoning and planning, and 25% fast and lightweight. In the same month, 63% of paying users used auto model selection at least once. The Project HydraFusion preview routed models per step (plan, implement, check) across almost 60,000 sessions. It is now generally available as GitHub Copilot Fusion.

Languages. From September 2025 to September 2026, TypeScript added 3.62 million monthly contributors, Python 3.23 million and JavaScript 1.82 million. Rust doubled. Merged PRs with detected AI involvement run about 40% to 44% for TypeScript, Python and Rust, and about 21% to 34% for Java, C# and C++. TypeScript was the primary language in 35% of 2026 repos with an agent signal, against 24% without one. For older repos the split was 27% against 17%. GitHub reads this as agents favouring a narrower set of languages when developers do not specify one. GitHub forecasts TypeScript and Python near 13 million monthly contributors each by September 2027, and JavaScript at 7.4 million.

Industries. A language model classified random samples of public repos created from 2022 to 2026 into NAICS sectors. It used each repo's name, description, topics and README. Administrative services, mining, health care and agriculture show over 40% of repos mentioning AI but under 3% with agent-opened PRs. Construction has about 25% AI mentions and about 9% agent-opened PRs, the highest of any sector, with manufacturing next. Professional, scientific and technical services grew about 5x since 2022 and rose from fourth to second place. Mining grew about 65x, but from a small base.

Platform and security. GitHub Actions ran 124 million jobs a day, more than three times last year. Developers merged over 30 million Dependabot PRs, about double 2025. Push protection ran 5.8 billion secret scans, nearly four times 2025's pace. Use of AI-assisted code review and Advanced Security each grew over 90%. Stacked pull requests entered preview on July 30, 2026 and had 264,000 users by September. Public repos mentioning interoperability or reverse engineering rose from about 30,000 to over 130,000.

Developer growth. The United States (38.3 million) and India (32 million) each added more than 10 million developers, and Brazil added 3.28 million. Against last year's 2030 projections, the US has closed 38.6% of the gap and India 28.3%. GitHub ran more than 40 forecasting models to 2031. About half predict more than 3x growth, just under half predict 2x to 3x, and three predict under 2x.

Limits

All figures are GitHub's own telemetry. The page does not explain how "detected agent involvement" or an "agent signal" is identified. The 8x commit gap and the TypeScript skew are correlations, and GitHub's reading of them is its own interpretation. The industry figures rest on LLM classification of samples whose size and accuracy the page does not report. Model-share figures exclude auto selection, bring-your-own-key requests, staff usage and monitoring traffic, and the page names no individual models. The three model-selection "patterns" and the regional breakdowns are referenced but missing from the fetched text. GitHub says its own forecasting models no longer agree on growth to 2031.