On 15 October 2026, GitHub announced support for several repositories within one Slack or Microsoft Teams conversation. A repository stores a project’s code.
For a small development team, this offers a way to retain shared context. Each coding task still needs a clearly defined scope.
What changed
GitHub’s examples include creating issues, listing open pull requests and investigating a bug across repositories. A pull request, or PR, proposes code changes for review and incorporation. The company separately says Slack can use shared messages and threads as context.
Copilot cloud agent is an AI assistant that carries out coding tasks in a cloud environment. Its current documentation limits each run to changes in its specified repository, one branch at a time and at most one PR per task. A shared discussion about several parts of a product therefore needs separately planned changes for each part.
Who can use it
The Slack and Teams guides label the integrations public previews. They require a paid Copilot plan, enabled cloud sandboxes (isolated execution environments) and appropriate repository access; organisation policies may restrict use.
Triggering code changes requires write access. Workspace guests and outside repository collaborators cannot start or steer these sessions. If a business works with an external contractor, check their actual role and access first.
Check the conversation before starting
GitHub warns that the agent captures the entire thread as context; that context is stored in the artifacts it creates. Direct-message actions use the linked personal GitHub account’s permissions; artifacts in shared conversations are created under the app’s identity.
Our editorial recommendation is to start a short, separate thread without customer data, secrets or unrelated agreements. Check who will be able to read the resulting issues and PRs.
How to try it on one defect
This is an editorial proposal, not a workflow we have tested:
- Choose a reproducible bug in a test environment, such as an order-status mismatch between the storefront and backend. Supply only authorised test data.
- Ask the developer to investigate the cause first and split the work by repository. Explicitly name the repository and base branch in every change request, rather than relying on channel settings.
- Agree acceptable spending and an accountable reviewer for each PR. Check the code and the interaction between product components before release. Our article on human review of AI output can help clarify responsibility.
The practical takeaway: the update brings discussions together, while the team needs to coordinate changes, checks and release order separately. Nadali has not tested these integrations in a working account or measured time savings.
Author: Nadali. Prepared using AI. Sources checked on 19 October 2026.
