GitHub used its 2026 Universe event to set out a vision for an end-to-end software delivery system built for developers working alongside AI agents. The company said the focus is shifting from producing code to completing software work, including features, bug fixes and security fixes.

GitHub described software development as a continuous process rather than a linear sequence of planning, coding, testing and deployment. Developers would set direction and remain accountable for what ships, while agents could implement and verify changes. GitHub said the system is intended to improve speed, use shared context to learn from work, and help catch security issues and bugs before release.

Among the products announced was the GitHub Copilot app, which connects agents and tools across projects. GitHub said users can direct work through the app and use it for tasks beyond coding, including writing specifications, discussing designs and analyzing telemetry.

Copilot Fusion is designed to select and coordinate one or more AI models based on a task’s needs, such as reasoning, code generation, debugging and tool use. GitHub cited results from the Artificial Analysis Coding Agent Index, where Fusion Max scored within two points of Opus 5.5 Max while costing half as much and taking 40% less time per task.

Copilot projects, available natively in the Copilot app from the day of the announcement, let users run agent work across multiple repositories in parallel managed sessions. GitHub also introduced session continuity, which lets users start and follow the same cloud agent session across the GitHub app, VS Code, the command-line interface and the web.

Copilot automations can run in the cloud manually, on a schedule or in response to an event. GitHub said they can coordinate agents for work such as issue triage and code review, and can now use outcomes to suggest improvements for users to review and approve.

For code review, GitHub introduced Copilot code review Max, a deeper review option for changes that need additional scrutiny. The company said Max recorded the highest F1 score on ReviewBench for high-severity issues; Lite and Balanced remain options for faster everyday reviews.

GitHub Enterprise customers can use Copilot features, including the app, code review and automations, without a separate Copilot subscription, paying instead for the AI credits they use. The option is available through GitHub Sales, with self-service access planned for later. Existing Copilot Business and Enterprise customers can keep their plans or switch when ready.

GitHub also outlined infrastructure work to support the changing demands of agent-assisted development. It said it has added four million CPU cores, 160 petabytes of high-speed storage and 50 terabits per second of network capacity so far this year. The company is migrating to a multi-region deployment in Azure and expects to complete that work early next year.

The company is also changing its Git architecture, guided by plans to simplify primitives, separate storage from compute and minimize coordination. GitHub said the work forms part of broader efforts across its AI gateway and agent harness to improve the platform’s capacity and reliability.

Sources: GitHub Blog.