News

GitHub Previews Agentic AI in VS Code Copilot

• By David Ramel
• 02/06/2025

GitHub today announced a raft of improvements to its Copilot AI in the Visual Studio Code editor, including a new "agent mode" in preview that lets developers use the AI technology to write code faster and more accurately.

Those improvements range from other previews, to new AI models to the general availability of Copilot Edits, which allows developers to make iterative changes across multiple files using natural language prompts (see "Blitz of New VS Code AI Improvements Include 'Copilot Edits'" and "Copilot AI Boosts Abound in New VS Code v1.96") about Copilot Edits in recent updates to VS Code. Along with the GA in VS Code, Copilot Edits is now in preview for Visual Studio 2022.

Those other previews as presented by GitHub include:

• Next edit suggestions to accelerate code changes by automatically identifying and proposing the next edit based on the context of previous changes. By simply pressing tab, users can instantly implement suggestions throughout an open file with insertions, deletions, and replacements.
• Prompt files allow users to store and share reusable prompt instructions in their VS Code workspace. These "blueprints” include self-contained markdown files that blend natural language guidance, file references, and linked snippets to supercharge coding tasks.
• Vision for Copilot enables users to bring a mock up to life by simply feeding Copilot a snip, screenshot, or image. From there, Copilot generates the UI, alt text, and code to go from vision to reality in minutes.
• New models from industry leaders Google's Gemini 2.0 Flash and OpenAI's o3-mini are now available in public preview in Copilot Chat with organization-wide access control to give administrators more choice over which models they want their developers to build with.

GitHub, however, emphasized the new agent mode to further agentic AI (a blog post is headlined "GitHub Copilot: The agent awakens"), probably the hottest area of advanced AI development right now. It represents a significant leap forward in GenAI, moving beyond passive assistance to proactive problem-solving. Agentic AI systems, instead of simply reacting to user input, are designed to perceive their environment, make decisions, and take actions to achieve specific goals, much like a human would. In addition to reacting to given instructions they're capable of figuring out how to accomplish objectives.

"GitHub Copilot's new agent mode is capable of iterating on its own code, recognizing errors, and fixing them automatically," GitHub said. "It can suggest terminal commands and ask you to execute them. It also analyzes run-time errors with self-healing capabilities.

"In agent mode, Copilot will iterate on not just its own output, but the result of that output. And it will iterate until it has completed all the subtasks required to complete your prompt. Instead of performing just the task you requested, Copilot now has the ability to infer additional tasks that were not specified, but are also necessary for the primary request to work. Even better, it can catch its own errors, freeing you up from having to copy/paste from the terminal back into chat."

To try it out, users need to enable the mode in a VS Code Insiders download; just look for "copilot agent" in VS Code's settings.

Furthering the company's agentic AI initiatives, GitHub also announced Project Padawan, providing SWE agents, AI-driven systems designed to assist software engineers by automating various development tasks. They can generate and review code, refactor and optimize codebases, automate workflows (such as tests and CI/CD pipelines), and provide architectural guidance, error troubleshooting, and best practices. These agents help developers by handling routine or specialized tasks, allowing engineers to focus on more complex, high-value work.

"We believe the end-state of Project Padawan will result in transforming how teams manage critical-yet-mundane tasks, such as fixing bugs or creating and maintaining automated tests," GitHub said.

GitHub also announced provisioning and authentication support for Copilot Workspace for Enterprise Managed Users, helping organizations to securely configure and control Workspace access. Developer teams can now use Copilot Workspace's agentic capabilities to quickly go from brainstorming to functional code. GitHub said Copilot Workspace leverages a system of sub-agents to iterate with developers and streamline collaboration across teams, helping to generate a plan, implement code, or automatically find and fix errors.

That again involves agentic AI, and that's clearly where the company is focusing its Copilot improvement efforts. "Developer teams will soon be joined by teams of intelligent, increasingly advanced AI agents that act as peer-programmers for everyday tasks," said GitHub CEO Thomas Dohmke in a news release. "With today's launch of GitHub Copilot agent mode, developers can generate, refactor and deploy code across the files of any organization's codebase with a single prompt command. Organizations who empower their developers to work with these agents will exist in another spectrum of productivity entirely." Dohmke also authored the blog post.

Here's a video to see agent mode in action.