A product decision may begin in a thread. Someone shares a file, another teammate points to an image, and the group agrees on what to change. By the time someone opens an issue or asks an agent to write code, part of that reasoning can be missing. The team then has to explain the work all over again.
In a September 25 update, GitHub said Copilot in Slack and Microsoft Teams can carry more of that context into development work. In Slack, supported files, attachments and message links can inform a request. In Teams, the agent can use inline images, forwarded messages, and channel or thread history. GitHub says Copilot also checks for similar issues before creating a new one, links to the resulting work and keeps a link back to the original conversation.
The value lies in keeping the decision trace when work leaves the chat.
A pull request linked to its discussion gives reviewers a chance to understand why a change was requested without reconstructing the story from memory.
Salesforce was already connecting conversation and action
There is useful precedent here. Salesforce describes Agentforce using conversations in Slack channels, lists and canvases to inform business actions, including opening a Service Cloud ticket. Slack Code already brings discussions, code diffs and previews into shared code channels, where teams can call agents such as GitHub Copilot. GitHub's September update improves what its own Copilot integration can carry from Slack and Teams into GitHub issues and pull requests; it does not introduce the broader idea of agents working inside chat.
Agentforce Vibes addresses a related but distinct step. Salesforce says its development agent understands a Salesforce project's schema, metadata and code to help plan, build, test and deploy apps.
A Salesforce project's context is not automatically the history of a Slack thread.
The documentation reviewed does not establish a direct equivalent of GitHub's conversation-to-issue-and-PR workflow. Teams using both platforms should ask which context reaches each agent, where decisions remain visible and who reviews the output.
Decide what should travel with the request
More context deserves a closer look. GitHub's Slack documentation says the cloud agent captures the whole thread as context and that this context is stored in generated artifacts. In a shared thread, a user with repository write access can start a task while messages from other participants can steer it. Where the request begins therefore matters as much as how it is worded.
A practical question for teams is whether everything in that thread should appear in the resulting issue or pull request. If the answer is unclear, a direct message or a narrower conversation may be a better starting point. Identity also changes with the setting: GitHub says the agent uses the linked personal account in Slack direct messages and its app identity in shared contexts.
Better controls, still a preview
The update also lets users change the model for their next message and retain that choice through the conversation. Slack users can set default owners and repositories. According to GitHub, the update also addresses longer tasks, interrupted replies, repository switching and Teams thread history. Teams will need to assess how reliably those changes work in their own processes.
The integration remains in public preview for organizations on Copilot Business or Enterprise. Usage draws on existing cloud agent entitlements and budgets, while some capabilities are rolling out gradually. Before wider adoption, a team could try one real workflow: a thread with an attachment, a possible duplicate issue and a pull request based on the discussion. Success would mean more than a completed task. Reviewers should be able to follow the decision trail and see when the agent carried more context than the work needed.
Sources: GitHub's September 25 update, integration documentation, Salesforce on Agentforce in Slack, Slack Code, and Salesforce on Agentforce Vibes.