GitHub announced on October 15 that software teams can now ask Copilot about several code repositories in a single Slack or Microsoft Teams conversation.
A repository is a project's code store. GitHub's examples include reviewing proposed code changes, known as pull requests, across projects; summarizing recent activity; and investigating a problem spanning a website's interface and its server-side code. Those questions can stay in one discussion instead of separate conversations for each repository.
Broader questions, not a promise of coordinated fixes
The practical distinction is between investigating a problem across projects and asking an agent to fix every affected project. The announcement establishes cross-repository questions, not coordinated editing across multiple repositories. It does not specify a repository-count limit or detailed read-only access rules for the new questions.
Triggering code changes still requires write access—permission to modify the repository—even though other participants can contribute context. The Slack guide and Teams guide also exclude workspace guests and repository outside collaborators from starting or steering sessions.
A useful first task is a cross-project progress question whose answer can be checked against the underlying pull requests. GitHub's agent guidance warns that responses can be inaccurate or incomplete; a summary should inform a team's next step, not replace review of the work.
Check access and choose the thread
The current Slack and Teams guides describe a public preview requiring a paid Copilot plan, the relevant GitHub app, a linked GitHub account and enabled cloud sandboxes. Organization or enterprise owners may need to enable access.
That is broader eligibility wording than GitHub's September 25 announcement, which named Business and Enterprise organizations and a gradual rollout. Follow the current setup guides, but check availability in your workspace before planning around it.
Choose the conversation deliberately. GitHub says the cloud agent captures the entire thread and stores that context in the work it generates; the integration guidance suggests a direct message to limit context. A focused thread or direct message lets a team supply the discussion relevant to the task without bringing unrelated conversation along.
The October update also describes better use of shared Slack messages and threads as context; that particular improvement is specified for Slack.
Two costs to check
Copilot chat and cloud-agent work consume AI credits, each worth US$0.01. The cost depends on the model and how much text it processes and generates, not a fixed charge per question. For organizations and enterprises, paid usage beyond included credits is enabled by default; an administrator can disable it.
When work runs in a cloud sandbox—the hosted environment for an agent task—compute, allocated memory and retained snapshots are metered separately. An AI-credit budget does not cap sandbox charges. Teams enabling agent work should therefore set both model-usage and sandbox spending controls, rather than assuming one budget covers the whole conversation.