GitHub Copilot app lets developers manage multiple AI agent sessions in one workspace

GitHub has released the Copilot app, designed to reshape how developers interact with AI coding assistance by moving beyond a single chat window into a multi-session workspace anchored to projects. The app recognizes that real software development is non-linear: developers often jump between bug fixes, code reviews, and exploration rather than following a straight path.

The core structure centers each session on a project. When you select a project from the Copilot app home screen (either from a previous project or by adding one from GitHub or your local machine), you gain an agent session with immediate access to the codebase, files, and tools for that task. For example, if you want to add a breadcrumb navigation component to an existing application, you can describe the update, and the session examines the project, identifies relevant files, makes changes, and runs tests—eliminating manual file searching and environment setup.

Once a session is underway, you can create additional sessions for other work without interrupting the original. Quick Chat provides a lightweight way to start a new conversation from the home screen—you might use it to ask about Copilot functionality (such as how worktrees operate) or investigate how your codebase works while another session continues on a separate task. When you return to your original session, you pick up where you left off, review changes, and move forward.

For UI work, the app introduces browser canvas, an interactive space that displays your application directly within the workflow. Rather than opening a terminal and separate browser, you use the /create-canvas slash command to preview your application. If a UI component needs adjustment, you can enable Canvas Dev Mode and use Pick & Polish to point directly at elements on the page and request changes, with each refinement informed by the visual result.

Once changes are ready, Agent Merge extends the workflow into the pull request and review phase. After you open the pull request options in the Copilot app and select Agent Merge, you can configure which actions it takes—addressing review feedback, resolving CI failures, or handling merge conflicts. Agent Merge then monitors the pull request as it moves through review and CI, assisting with any issues that arise, and preparing the pull request for merge once required checks pass.

The GitHub Copilot app addresses a fundamental mismatch between how AI chat tools are typically designed and how developers actually work. Traditional chat-based AI coding assistants assume a linear, question-and-answer workflow, but real software development is context-heavy and task-switching is constant. A developer might begin implementing a feature, discover they need to understand an unfamiliar part of the codebase, pause to review a colleague's pull request, and return to the original task—each step requiring different context and tooling.

By anchoring agent sessions to projects and allowing multiple concurrent sessions, the Copilot app preserves context across these transitions. The addition of browser canvas—letting developers visualize UI changes directly within the workflow—bridges the gap between code review and visual feedback, a gap that normally requires opening a separate terminal and browser. Agent Merge's automation of pull request workflows (addressing feedback, handling CI failures, managing merge conflicts) extends the AI's role beyond initial code generation into the full development lifecycle, reducing friction at a stage where many code changes currently stall.

No comments yet. Be the first to share your thoughts!