Microsoft has introduced a new feature for GitHub Copilot today: the ability to perform tasks or solve problems like a developer while running in the background, functioning as an artificial intelligence agent.

"This agent launches a secure and fully customizable development environment powered by GitHub Actions," said Thomas Dohmke, CEO of GitHub under Microsoft. "As the agent works, it pushes commits to draft pull requests, and you can track every step through the agent session logs."

GitHub Copilot can now act as an asynchronous AI development partner, directly integrated into the GitHub platform.

This new feature incorporates agent-based AI, a growing trend in artificial intelligence that enables models to complete tasks autonomously and work towards goals with minimal human supervision in most scenarios. Although the new GitHub Copilot still requires human oversight—its pull requests need approval and run continuous integration and deployment workflows—it shows remarkable capability in well-tested codebases for tedious tasks and low to medium complexity jobs such as adding features, fixing bugs, extending tests, refactoring code, and improving documentation.

"The GitHub Copilot coding agent is opening doors for every developer to have their own team, working in parallel to enhance their capabilities," said James Zabinski, DevEx Lead at Ernst & Young Global Ltd.

Microsoft stated that the agent is powered by the Model Context Protocol (MCP), a toolkit that connects AI models to external data and functionalities outside GitHub. The MCP server can be configured in the repository settings. The agent also supports multimodal capabilities, allowing it to view images included in assigned GitHub issues, making it possible to develop new features using error screenshots or model diagrams.

"Whether it's code completion, suggesting the next edit, chatting, agent mode, or now the coding agent, GitHub Copilot always has one mission: keeping you in that magical flow state," Dohmke added.

GitHub Models Allow Developers to Experiment

As developers collaborate with AI and Copilot, the new GitHub Model Hub will allow users to explore top-tier models to create, store, evaluate, and share prompts without leaving GitHub, Microsoft noted.

GitHub Models will serve as a centralized hub for model and prompt evaluations, enabling users to build, test, and manage AI functionalities directly from their environments and repositories, eliminating the need to switch contexts between tools.

Microsoft also mentioned that it will allow developers to experiment with safeguards so they can operate securely.

Introducing Agent-Based AI to DevOps

Today, Microsoft also announced the launch of what it calls Agent-Based DevOps, the next evolution of DevOps, where intelligent agents will collaborate with users and each other. These agents will automate and optimize every stage of the software lifecycle.

DevOps is a software development methodology that combines 'development' and 'operations' teams to collaboratively take responsibility across the software lifecycle, integrating code into repositories and continuously testing and delivering software through automation. It also includes continuously measuring application performance.

"Agent-Based DevOps will help you build faster, eliminate backlogs, reduce technical debt, secure your applications, and keep everything running smoothly in production," Microsoft explained. "The best part is, we place you at the center of this orchestra, directing the agents and approving suggestions so you can focus on building epic solutions."

The company says GitHub Copilot will be at the heart of all this, capable of handling complex multi-step coding tasks, analyzing intricate codebases, making cross-file edits, generating and running tests, fixing bugs, and suggesting commands.

Microsoft also announced the release of a Site Reliability Engineering agent for Azure, which can run around the clock and automatically troubleshoot issues as they arise. It continuously monitors the performance of Kubernetes, app services, serverless, and database environments in real-time, leveraging deep knowledge built from Azure-based services.

When it provides remedies or fixes, these actions are logged in GitHub issues for teams to follow up and close the loop. This means fewer emergency calls, as the system can self-heal. Of course, it can't fix everything on its own—someone may still get woken up.

Open-Sourcing GitHub Copilot in VS Code

Microsoft is open-sourcing GitHub Copilot in Visual Studio Code. Over the next few months, the company plans to bring the AI-powered features of the GitHub Copilot extension to the VS Code open-source repository.

VS