GitHub unveils Copilot agent that writes and fixes code autonomously

GitHub’s new AI coding agent represents a significant evolution in developer tools, automating complex coding tasks directly within the popular Copilot platform. This advancement positions GitHub alongside competitors like Google and OpenAI who have recently unveiled similar AI coding agents, signaling a major shift in how software development workflows are being reimagined through AI automation.

The big picture: GitHub has integrated an AI coding agent into Copilot that can autonomously fix bugs, add features, and improve documentation after being assigned a task.

How it works: The AI coding agent operates by automatically spinning up a virtual machine, cloning the repository, and analyzing the codebase to complete assigned tasks.

• The agent saves changes as it works and provides detailed reasoning in session logs.
• When finished, it tags the developer for review and can automatically address any feedback comments.
• The system incorporates context from issue discussions and pull request conversations while following repository-specific coding standards.

Competitive landscape: GitHub’s move follows similar recent announcements from other major AI companies developing autonomous coding agents.

• Google unveiled Jules in December 2024.
• OpenAI introduced ChatGPT’s coding agent, Codex, just last week.

Availability details: The new coding agent is accessible to users of Copilot Enterprise and Copilot Plus through multiple platforms.

• The agent can be accessed via GitHub’s website, mobile app, and Command Line Interface tool.
• Microsoft has also announced it’s open-sourcing GitHub Copilot in Visual Studio Code, allowing developers to build upon the tool’s AI capabilities.