GitHub has recently unveiled a suite of new features for GitHub Copilot, aimed at simplifying coding tasks and boosting productivity for both developers and organizations alike.

The standout update is the general availability of Copilot Edits, a feature that empowers users to identify a collection of files for editing while using natural language to instruct Copilot to implement inline changes across multiple files swiftly. This enhancement optimizes workflow within a user’s environment.

Several new preview features are now available in Visual Studio Code. Notably, the Agent mode feature enables Copilot to revise its outputs autonomously, ensuring that the results meet the user’s requests fully.

Additionally, Copilot can now automatically detect and rectify errors, suggest terminal commands, and analyze run-time errors with impressive self-healing abilities. The next edit suggestions function allows for the automatic identification of logical code edits, which users can implement simply by pressing the tab key.

Improvements to prompt files enable developers to store and share reusable prompt instructions within their VS Code workspace. These “blueprints” aim to streamline coding tasks by combining natural language guidance with file references and linked snippets. The Vision for Copilot introduces another exciting feature, allowing users to convert mock-ups into reality by submitting images or screenshots, prompting Copilot to generate user interfaces, alt text, and the necessary code.

Furthermore, Copilot Chat now includes new models from Google’s Gemini 2.0 Flash and OpenAI’s o3-mini in public preview, along with organization-wide access controls, allowing admins to regulate which models developers utilize for their projects.

GitHub’s CEO Thomas Dohmke emphasized the transformative role of AI agents in development teams, asserting that intelligent AI agents will soon act as collaborative peers for daily tasks. He stated, “With today’s launch of GitHub Copilot agent mode, developers can generate, refactor, and deploy code across any organization’s codebase with a single command.”