GitHub Octoverse 2026: AI Agents Now Initiate Over 40% of Pull Requests

The 2026 GitHub Octoverse report reveals an astonishing surge in platform activity, setting new benchmarks for developer engagement. In the 12 months leading up to September 30, developers across the globe made a staggering 17.9 billion pushes to GitHub repositories, a figure that surpasses the combined totals of 2024 and 2025. This exponential growth isn't limited to pushes; commits across both public and private repositories soared to 34.3 billion, more than tripling the volume observed just two years prior. Merged pull requests also experienced a dramatic increase, reaching 1.3 billion over the same period, averaging an impressive 3.5 million per day.

This widespread increase in activity indicates a thriving and expanding developer community, leveraging GitHub as the central hub for their projects. The report highlights that this growth spans every industry and field, from critical infrastructure projects like Nepal's Gati flood bulletin, built on GitHub Pages and Actions, to cutting-edge software for spacecraft developed at NASA's Jet Propulsion Laboratory. Furthermore, public contributions alone more than doubled in 2026, reaching 2.5 billion, compared to 1.12 billion in the previous year. This sustained momentum suggests a robust and increasingly collaborative ecosystem, driven by both individual curiosity and collective innovation.

Perhaps the most striking revelation from the 2026 Octoverse report is the transformative role of AI agents in the development lifecycle. GitHub's data indicates a significant shift where AI agents are no longer just assisting but are actively initiating a substantial portion of development work. As of September 2026, agents initiated over 110 million pull requests, accounting for more than 40% of the month's total. This marks a dramatic increase from just 17% in March of the same year, demonstrating a rapid adoption curve for AI-driven code generation and integration.

The report suggests that for a growing segment of development tasks, the initial step now involves directing an AI agent, with human developers then focusing on reviewing and refining the output. This paradigm shift is further supported by observations that repositories with agent configuration files averaged eight times as many commits as new repositories without them, indicating a strong correlation between AI integration and higher development velocity. Developers are also demonstrating flexibility in their AI toolchains, with 61% of GitHub Copilot users leveraging more than one model in September, spreading their requests across general-purpose, deep reasoning, planning, and fast, lightweight models. This highlights a growing sophistication in how developers interact with and orchestrate various AI capabilities to suit different aspects of their projects.

The increasing prominence of AI agents is not just about raw code generation; it's fundamentally altering developer workflows and fostering new forms of collaboration. With AI handling the initial scaffolding or repetitive tasks, developers can dedicate more time to complex problem-solving, architectural design, and thorough code review. The report notes that review is becoming the critical juncture where developers shape the AI-generated output. This emphasis on review is supported by the introduction and rapid adoption of tools like stacked pull requests, which launched as a preview on July 30 and were used by over 264,000 developers by September, facilitating easier sequential review of smaller, related changes.

Furthermore, the usage of AI-assisted code review and GitHub Advanced Security both saw over 90% year-over-year growth, indicating a holistic approach to integrating AI not just for creation but also for quality assurance and security. This suggests a future where AI acts as an intelligent partner throughout the entire development pipeline, enhancing both productivity and code integrity. The report also highlights that developers are working across 10% more repositories on average compared to a year ago, showcasing an increased breadth of engagement and collaboration within the GitHub ecosystem.

The 2026 Octoverse report provides a compelling glimpse into the future of software development, a future increasingly shaped by intelligent agents. While the data clearly shows a strong correlation between AI adoption and increased activity, GitHub emphasizes that these are platform measurements and do not inherently prove that agent-written changes are more reliable. However, the sheer volume of agent-initiated pull requests signals a permanent shift in how code is conceived and constructed. The trend towards multi-model usage among Copilot users also indicates a growing sophistication in AI-assisted development, moving beyond single-tool reliance to a more integrated, intelligent environment.

As AI capabilities continue to advance, developers will likely see even more specialized agents for various tasks, from testing and debugging to deployment and infrastructure management. The focus for human developers will increasingly shift towards higher-level design, strategic decision-making, and critical oversight of AI-driven processes. The Octoverse 2026 report serves as a powerful reminder for individuals and organizations to embrace these changes, invest in AI literacy, and adapt their practices to thrive in this new era of agent-assisted software engineering. The momentum is undeniable, and the developer community is clearly leaning into this transformative technology at an unprecedented pace.