GitHub Launches New Copilot AI Agent That Could Replace Your Late-Night Debugging
A powerful new AI coding agent from GitHub automates tasks in the cloud to boost developer productivity.
Key Takeaways:
- GitHub unveils a new Copilot AI agent that works asynchronously in the cloud.
- The agent can handle coding tasks like bug fixes and documentation updates.
- It’s available to select users and brings major upgrades to developer workflows.
Microsoft’s GitHub has introduced a new Copilot AI coding agent that operates asynchronously within GitHub Actions, tackling tasks in the background. This autonomous assistant can seamlessly add new features, fix bugs, and enhance documentation—all without interrupting the developer’s workflow.
Microsoft first announced the GitHub Copilot Agent as Project Padawan back in February. It’s different from the GitHub Copilot agent mode that debuted in February to support synchronous (real-time) collaboration. Agent mode works within the users’ development environment on their computer.
How does the new Copilot AI Agent work?
This new Copilot agent is designed to operate entirely in the cloud. It leverages GitHub Actions to establish a secure, customizable development environment on demand. This capability allows the coding agent to work asynchronously, which can handle tasks in the background while developers focus on their work. When invoked, the AI coding agent can navigate the repository, edit files, run commands, and open pull requests.
“It uses GitHub Actions behind the scenes to boot a virtual machine, clones the repository, configures the environment and analyzes the codebase with advanced retrieval-augmented generation (RAG) [in] GitHub code search,” GitHub explained. “As the agent works, it regularly pushes its changes to a draft pull request as git commits and updates the pull request’s description. Along the way, you can see the agent’s reasoning and validation steps in the session logs.
Integration with Model Context Protocol
Microsoft mentioned that the AI coding agent can also be configured to work with model context protocol (MCP) servers to fetch missing or external information it needs to complete a task. For instance, the agent is working on a GitHub issue that involves fixing a broken image in a web app. The code might reference an image that no longer exists or is corrupted. The agent contacts the MCP server of the relevant data source to get the missing image or metadata.
The new coding agent (which costs about $400/year) is currently available for Copilot Enterprise and Copilot Pro Plus customers via GitHub’s official website, the GitHub mobile app, and the GitHub Command Line Interface tool.
In related news, Microsoft has announced that it’s open-sourcing GitHub Copilot in Visual Studio Code. This release will allow developers to view, modify, and contribute to the code. Microsoft has also announced a new feature called GitHub Models. It will allow developers to evaluate and compare models from different providers like OpenAI, Meta, Microsoft, Mistral, Cohere, and others.
Rabia has a master's degree in Software Engineering and she has years of experience writing professionally about Microsoft products and other technologies. Rabia has also written for OnMSFT.com as well as Windows Report. She is always up to date on t...