GitHub is set to introduce a new AI coding agent designed to assist developers in fixing bugs, adding features, and enhancing documentation without direct input. This agent will be integrated into GitHub Copilot and is activated when a user assigns a task, according to an announcement made during Microsoft Build.
To perform its functions, the AI coding agent creates a virtual machine, clones the relevant repository, and evaluates the codebase. It keeps track of its modifications and provides detailed session logs to explain its decision-making process. Upon completion, the agent will notify developers for review, allowing them to leave feedback that the agent can then address automatically.
In addition to GitHub’s initiative, other technology firms have also showcased their own AI coding tools. Google unveiled a coding agent named Jules in December, and recently, OpenAI presented its coding agent Codex, which operates through ChatGPT.
According to GitHub, the coding agent can utilize context from related issue or pull request discussions and adhere to specific repository guidelines, which helps it grasp the intent behind tasks and align with project coding standards. The new feature is accessible to users of Copilot Enterprise and Copilot Pro Plus via GitHub’s website, mobile application, and the GitHub Command Line Interface tool.
Additionally, Microsoft announced the open-source release of GitHub Copilot for Visual Studio Code, opening possibilities for developers to enhance the tool’s AI features.
Correction, May 19th: An earlier report inaccurately referred to the AI coding agent as being available through Copilot Plus; the correct term is Copilot Pro Plus.
