GitHub: Cloud agent sessions now available directly from issues and project views

GitHub announced a new feature on April 23, 2026, that connects Copilot cloud agents with the main task management interface. Developers can now track and manage agent sessions directly from issues and projects views.

What are cloud agent sessions and how did they work before?

Cloud agent sessions are asynchronous task executions in the Copilot ecosystem. Unlike the interactive Copilot in the editor, the cloud agent runs in a remote container — it receives a task description, explores the repository, modifies code, and opens a pull request, all without the developer being present.

Until now, tracking those sessions required switching to a separate interface. If a developer started the agent from an issue comment, they had to leave the context to see what the agent was doing. This fragmented the workflow, especially in teams that use GitHub Issues as the primary source of truth for planning.

What do session pills and side panels bring?

The new version introduces session pills — small visual indicators attached to issues that show whether an agent session is active or completed. Clicking the pill opens a side panel with real-time progress logs: which files the agent is reading, which commands it is executing, what decisions it is making.

In project views (GitHub Projects v2 kanban boards), sessions can be activated automatically based on rules. For example, when an issue moves to the “In Progress” column, the system can automatically start the agent for preliminary analysis. This turns the project board into a place where human and AI work combine seamlessly.

Who benefits most?

Teams that already centralize their work around GitHub Issues gain the most. A manager can see the status of the entire sprint along with AI agent progress, without jumping between tools. Junior developers gain transparency into how the agent solves a task, which accelerates learning.

This move is part of GitHub’s broader “agentic developer experience” strategy — gradually embedding Copilot functionality into native GitHub interfaces. The next logical steps are integration with Actions workflows and deeper connectivity with the code review process.