A typical development day begins in Visual Studio Code with a new task in the backlog.

“Add authentication to the application.”

The objective is clear. The technical path is familiar. What varies is the way the work is executed —whether through careful exploration, uninterrupted execution, or collaborative review. Copilot Agents are designed to support these different working styles without changing the underlying goal.

One Goal, Flexible Execution

Copilot Agents do not alter what is being built. They influence how GitHub Copilot assists during the development process. The same intelligence is applied across multiple execution modes, allowing developers to choose the level of interaction, autonomy, and collaboration appropriate for the task at hand.

Local Agent: Interactive Development Inside the Editor

During the initial phase of a feature, understanding the codebase and design decisions is critical. The Local Agent operates directly within Visual Studio Code, providing context-aware assistance while keeping the developer in control. It reads project files, answers questions, suggests code changes, and prepares edits without applying them automatically.

This mode supports iterative exploration and decision-making, enabling developers to refine logic and implementation details incrementally. The Local Agent is well suited for scenarios that require close interaction and continuous feedback.

Background Agent: Autonomous Progress Without Disruption

Once the implementation approach is established, uninterrupted execution often becomes the priority. The Background Agent works autonomously on the developer’s machine, applying changes without requiring active interaction. Development continues in parallel while the agent completes the assigned task.

This mode is effective when progress is needed without disrupting ongoing work, allowing results to be reviewed once execution is complete.

Cloud Agent: Collaborative and Team-Ready Execution

As changes move toward completion, collaboration and visibility become essential. The Cloud Agent runs in a remote environment that integrates naturally with branch-based workflows and pull request processes. It enables AI-generated changes to be shared, reviewed, and discussed by the broader team. This approach ensures that Copilot’s contributions align with established collaboration and governance practices.

Regardless of the execution mode — local, background, or cloud — changes can be surfaced through pull requests. This maintains:

• Transparency into modifications
• Human review and approval
• Consistency with existing development workflows

AI-assisted changes follow the same review standards as human-authored code.

conclusion

GitHub Copilot Agents provide flexibility in how AI assistance is applied throughout the development lifecycle. By offering multiple execution modes, Copilot supports individual development, autonomous progress, and team collaboration—while remaining consistent with established GitHub workflows and review practices.