Copilot Chat gains more context from agent sessions

GitHub has improved the connection between Copilot Chat and Copilot cloud agent sessions on the web. Published on June 10, 2026, the update lets Chat display the status of active sessions, answer follow-up questions using completed session logs, and search previous agent work without forcing developers to reconstruct the context manually.

{getToc} $title={Table of Contents}

Copilot Chat gains more context from agent sessions

Copilot Chat can now follow work delegated to a Copilot cloud agent more closely. When a user asks Chat to start a session, create a pull request, or perform deep research on a repository, the conversation can reflect the status of that work while the session remains active.

Once the session is complete, users can continue the conversation instead of treating the result as an isolated task. Chat can use the session logs to explain what changed, what validation was performed, and why the agent made specific implementation decisions. It can also start another session using the existing conversation as context.

{getCard} $type={post}

Two new tools connect Chat with previous agent work

The first addition is Get agent logs. This tool lets Copilot Chat retrieve the logs from cloud agent work associated with a pull request. Developers can ask questions about the files that changed, the tests or checks that were completed, and the reasoning behind the resulting implementation.

The second addition is Session search. It can find and summarize previous agent sessions using details such as topic, title, or recency. This can help a developer locate an earlier task, recover its context, and continue working without searching manually through multiple pull requests or session pages.

Why this matters for frontend and template projects

Frontend repositories often contain more than visible page code. They may include component libraries, build tools, CSS processing, template files, deployment settings, accessibility checks, tests, and content-generation scripts. Understanding exactly what an agent changed across those areas can take longer than reviewing a simple code diff.

Access to session logs gives designers and developers another way to examine that work. A team could ask whether an agent modified responsive styles, validated a build, updated dependencies, changed a reusable component, or encountered a problem while preparing a pull request.

Session search can reduce repeated investigation

Searching previous sessions can be useful when a design or development issue returns later. Instead of asking Copilot to investigate the same component or repository behavior from the beginning, a user can look for the earlier session and review the work already completed.

This can improve continuity in longer projects where several agent sessions are used for research, implementation, testing, and refinement. It also makes the handoff between automated work and human review clearer because the conversation can retain more of the task history.

IMPORTANT: Session queries only include sessions started by the current user. The sessions must also be synced to the user's GitHub account, and organization policies can prevent local session data from being stored in the cloud.{alertWarning}

Daisuki's Take: What This Means for Web Designers

For web designers working with code repositories, the useful part of this update is not simply that Copilot remembers more information. It is that the path between assigning work, checking progress, reviewing the result, and asking for clarification becomes easier to follow.

We think this can be particularly helpful when an agent works on template systems, responsive components, build errors, or frontend maintenance tasks. Those changes often depend on decisions that are not immediately obvious from the final visual result or code diff.

The practical takeaway is to treat session history as supporting context, not as a replacement for review. Asking what changed and what was validated can make agent-assisted work easier to understand, but designers and developers should still inspect the resulting interface, test responsive behavior, and confirm that the implementation fits the project.

{getCard} $type={custom} $title={What's New?} $info={More design, AI tools, creative workflow, and visual editing news} $icon={}

Sources and Recommended Links

• Copilot Chat now sees your agent sessions | GitHub Changelog
• Managing agent sessions | GitHub Copilot Official Documentation