Enhanced Observability for Copilot Coding Agent Sessions

GitHub has shipped several improvements to the Copilot coding agent's session logging interface, giving developers better insight into what their AI agent is doing while working on delegated tasks.

What's Improved

Built-in Setup Steps Visibility Before Copilot begins working on a task, it performs setup operations like cloning your repository and starting the agent firewall. These steps now appear in the session logs with clear start and finish markers, eliminating the mystery of why Copilot appears idle during initialization.

Custom Setup Step Output Developers who customize their development environment using a copilot-setup-steps.yml file can now see the actual output from those setup steps directly in the session logs. This eliminates the need to dig into verbose GitHub Actions logs when debugging environment configuration issues.

Subagent Activity Tracking The Copilot agent often delegates work to specialized subagents for tasks like code analysis or research. The session logs now collapse subagent activity by default with a status display showing current work, allowing developers to expand details on demand. This keeps logs readable while preserving full transparency.

Developer Impact

These improvements address a key pain point: understanding exactly what happened during a long-running agent session. Developers can now verify their setup is correct, spot where time is being spent, and debug failures more quickly without context-switching to other tools.

Next Steps

Visit the Copilot coding agent documentation to learn more about session logs and agent customization options.