At Microsoft Build, GitHub unveiled a major upgrade to its Copilot platform: agent capabilities that promise to automate entire DevOps loops. This new iteration of Copilot is designed not just to assist with coding but to act as an active, intelligent collaborator throughout the software development lifecycle.
From Code Assistant to Agentic Collaborator
The highlight of the announcement is a powerful asynchronous Copilot agent now integrated directly into GitHub and available through VS Code. This AI-driven teammate can take on end-to-end tasks, from issue assignment to implementation, while maintaining transparency and control for human developers. GitHub is branding this evolution as part of a “powerful Agentic DevOps loop.”
Developers can assign issues directly to Copilot or communicate tasks via Copilot Chat. Once set into motion, the agent operates within GitHub’s robust control systems, contributing code via draft pull requests, which remain visible and reviewable through detailed agent session logs.
A Secure, Customizable Environment
Security remains a central concern when integrating AI into software pipelines. GitHub has addressed this by embedding the agent within a secure environment powered by GitHub Actions—its established CI/CD platform. With over 25,000 actions available and 40 million jobs run daily, Actions provide a trusted foundation for Copilot’s automated efforts.
Safeguards such as branch protections, limited internet access, and mandatory human approval before CI/CD execution help ensure secure and auditable workflows. This design allows enterprises to adopt Copilot agents without compromising compliance or control.
Beyond the Editor
Though Copilot started as an in-editor assistant, GitHub is extending its reach. The Copilot agent experience is expanding to JetBrains, Eclipse, and Xcode, enabling multi-platform support for teams across different environments. Furthermore, GitHub is open-sourcing Copilot Chat in VS Code, promoting transparency and encouraging community innovation.
For even greater flexibility, the Model Context Protocol (MCP) allows teams to extend the agent’s awareness by integrating it with external tools and datasets, configured directly from repository settings.
Industry Voices Weigh In
Industry leaders are already seeing real-world potential in GitHub’s new approach. Kate Holterhoff, Senior Analyst at RedMonk, emphasized the shift:
“With its autonomous coding agent, GitHub is looking to shift Copilot from an in-editor assistant to a genuine collaborator in the development process.”
This sentiment is echoed by GitHub CEO Thomas Dohmke, who positions the platform as the future home of AI-driven development:
“It’s vital that organisations and developers are ready to embrace these agents without compromising their security posture.”
Early adopters are also seeing tangible benefits. At Carvana, the agent is already increasing development speed:
“It converts specifications to production code in minutes,” said Alex Devkar, SVP of Engineering and Analytics.
And at EY, the tool is helping developers stay focused:
“We’re now able to assign tasks that would typically detract from deeper, more complex work,” added James Zabinski, DevEx Lead.
Available Now in Preview
GitHub is rolling out the Copilot agent to Copilot Enterprise and Copilot Pro+ users starting today, in preview. Using the agent will count against premium Copilot request quotas and GitHub Actions minutes.
As Copilot transitions from code completion tool to autonomous DevOps partner, it’s clear that GitHub is betting big on AI agents as the next evolution in software development.
