At Microsoft Build, GitHub announced a major leap forward for its AI-powered development assistant—Copilot—by introducing agent capabilities designed to automate and streamline DevOps workflows.

This new asynchronous coding agent can now take initiative when needed, acting as a collaborative teammate embedded directly within GitHub and accessible via Copilot Chat in VS Code. GitHub calls it a step toward enabling a “powerful Agentic DevOps loop,” giving developers a more intelligent, action-ready assistant in their toolchain.

A New Era of Autonomous Development

The Copilot agent isn’t just a fancy autocomplete anymore. With these enhancements, Copilot can now:

• Accept and act on GitHub Issues
• Write and push code via draft pull requests
• Receive developer feedback through the standard PR review process
• Maintain transparency with session logs and traceable activity

Expanded Access and Open Source Momentum

To go along with the agent rollout, GitHub is also:

• Open-sourcing Copilot Chat for VS Code
• Extending agent mode to JetBrains, Eclipse, and Xcode
• Updating GitHub Models, including support for xAI’s Grok 3

This makes the Copilot agent ecosystem more inclusive for developers outside the Visual Studio Code bubble, further widening its reach.

Built for the Enterprise, With Security in Mind

GitHub CEO Thomas Dohmke emphasized that these agents are designed for secure, transparent, and customizable collaboration.

“GitHub is where the world’s developers work on their projects. Now, it’s becoming the place where they collaborate with agents in a configurable, steerable, and verifiable way,” Dohmke said.

“The Copilot coding agent is the most enterprise-ready of its kind—amplifying human developers with trust by design.”

Security is clearly a priority. The agent operates within GitHub’s existing guardrails, such as:

• Branch protections
• Controlled internet access
• Human approvals before any code enters CI/CD pipelines

All changes proposed by the agent still require a developer’s review and approval, ensuring that humans remain in control.

Deep Integration with GitHub Actions

At the core of the Copilot agent’s environment is GitHub Actions, GitHub’s CI/CD platform with more than 25,000+ reusable actions in the marketplace. The agent uses this infrastructure to execute tasks in a secure, isolated, and configurable manner.

With 40 million+ jobs run daily on GitHub Actions, this platform is already battle-tested across open-source projects and enterprise-grade software development. Now, Copilot taps into this power to deliver code faster, safer, and with less manual overhead.

Smarter Agents with External Context

To make the agent even more capable, GitHub is introducing the Model Context Protocol (MCP). This feature allows organizations to connect the Copilot agent to external data sources and tools outside the GitHub ecosystem. Configuration is simple—teams can set up MCP servers in the repo settings to expand the agent’s operational context.

Early Feedback: It Works

GitHub has already begun testing the Copilot agent internally and with selected enterprise partners. The early results show strong performance for low- to medium-complexity coding tasks, especially in well-tested codebases.

“The Copilot coding agent fits into our existing workflow and converts specifications to production code in minutes,” said Alex Devkar, SVP of Engineering at Carvana.

“This increases our velocity and enables our team to focus on higher-level creative work.”

James Zabinski, DevEx Lead at EY, added:
“The agent lets developers assign routine tasks that would typically distract from deeper, strategic work. Now, teams can focus more on high-value coding challenges.”

Available Now (in Preview)

Starting today, Copilot Enterprise and Copilot Pro+ users can try the new agent in preview mode. Keep in mind that using the agent will consume Copilot premium request quotas and GitHub Actions minutes.

GitHub’s new agent capabilities mark a pivotal shift—from Copilot being a helpful code suggestion tool to becoming an active engineering collaborator. By automating repetitive development tasks and working within existing DevOps pipelines, this agent could significantly transform how teams build and ship software.

The age of AI-powered software engineering is no longer a concept—it’s writing your next pull request.