GitHub's agent push: what changes for your repos

GitHub's Octoverse report, published October 28, 2026, puts a number on how much agent work has moved into daily development. Developers pushed code to GitHub 17.9 billion times in the past 12 months, more than in 2024 and 2025 combined. Commits across public and private repos reached 34.3 billion, more than three times the level of two years earlier.

The shift shows up in pull requests. As of September, developers start more than 40% of pull requests with an agent, up from 17% in March. Among repos created this year, the average one with an agent configuration file logged eight times as many commits as the average new repo without one. That is a comparison of averages, not proof that the config file caused the difference.

Model choice is spreading too. In September, 61% of GitHub Copilot users worked with more than one model, spreading requests across general purpose, deep reasoning and fast lightweight models.

At GitHub Universe, GitHub's Jay Parikh described a shift from writing every line of code to orchestrating fleets of agents. The stated goal is completed work, not more code. GitHub framed the platform as a software delivery system where developers set direction and stay accountable, while agents implement and verify changes.

The product news centers on the GitHub Copilot app, built for agentic delivery. It connects agents and tools across projects and also handles specs, design discussions and telemetry analysis. Copilot projects, available natively in the app from that date, run agent work across multiple repos in managed parallel sessions.

Session continuity lets you follow the same cloud agent session across the GitHub app, VS Code, the CLI and the web. Copilot automations can now learn from their outcomes and propose improvements for you to review and approve.

Copilot Fusion picks how to execute a task across one or more models, based on what the task needs across reasoning, code generation, debugging and tool use. GitHub reports that Fusion Max scored within two points of Opus 5.5 Max on the Artificial Analysis Coding Agent Index, at half the cost and 40% less time per task. Those are GitHub's own figures from a named benchmark, not an independent test.

On review, GitHub introduced Copilot code review Max, which it says posted the highest F1 score on ReviewBench for high-severity issues. Lite and Balanced modes stay for everyday reviews. More code means more pull requests, so review capacity is the constraint GitHub is addressing.

For enterprise customers, all Copilot features are now usable without a separate subscription, paid through AI credits consumed. GitHub says this runs through GitHub Sales, with self-serve coming later. Existing Copilot Business and Enterprise customers can keep current plans or move when ready.

All of this depends on capacity. GitHub says it added 4 million CPU cores, 160 PB of high-speed storage and 50 Tb/s of network capacity so far this year. It is migrating to a multi-region deployment in Azure and expects to finish early next year.

On the Git side, GitHub's engineering post from October 21, 2026 describes a change to how writes are serialized. The quorum scope moves from the whole repository to individual references, so writes to different branches no longer wait on each other. In internal testing with realistic traffic, writes to different references saw 3–4x higher throughput at lower latency. That is internal testing, not a public benchmark.

GitHub also reports that 99.99999% of repositories now live in two places, its own data centers and Azure, with about half of Git traffic served from Azure. The remaining busiest repos were held back by a roughly 25ms cross-region round trip.

The numbers describe platform-wide activity, not your team. A repo with an agent config file logging more commits says nothing about whether those commits are good. GitHub's own framing keeps developers accountable for what ships, and review is where that happens.

The cost story has a catch. Enterprise access without a separate Copilot subscription still means paying for AI credits used, and self-serve availability is not yet there. If you are on Copilot Business or Enterprise, you can stay put until you decide.

What remains open: whether the throughput gains hold outside GitHub's internal tests, and when the Azure migration completes. GitHub says early next year. Until then, treat the performance claims as vendor-reported.