GitHub Copilot coding agent now works in Slack, letting you mention @GitHub in any thread with a prompt like “Fix missing tooltips on navigation icons,” and it will generate a pull request in the background.
The feature is straightforward: the GitHub app for Slack gets upgraded to support the Copilot coding agent. You prompt it, it works asynchronously, and replies in the thread when the PR is ready for review.
Why Slack matters#
The interesting part isn’t that Copilot can generate PRs—it’s been doing that for months. It’s that you can now trigger it from wherever your team already discusses bugs, features, and tech debt.
Slack is where someone says “users are complaining about missing tooltips” and someone else replies “I’ll look at it later.” Now you can just mention @GitHub right there and let the agent handle it.
This reduces friction. You don’t need to switch to your IDE, open the repo, draft a prompt in GitHub, and wait. You stay in the conversation, delegate the task, and come back when it’s done.
The async workflow#
Copilot coding agent is designed for asynchronous work: bug fixes, incremental features, test coverage, refactors. Tasks you’d normally put in your backlog and get to eventually.
The Slack integration reinforces that. You’re not blocking your flow waiting for code to generate. You’re offloading work that doesn’t need immediate attention, then reviewing it when you have time.
The question is whether that changes how teams prioritize work. If it’s easy to spin up an agent for every small fix, do you stop batching tasks? Does the backlog shrink or just turn into a queue of PRs waiting for review?
What you need#
To use it, you need Copilot coding agent enabled (available with all paid Copilot plans, but Business and Enterprise subscribers need admin approval), the GitHub app installed in Slack, and your GitHub account linked.
The integration is optional and requires new permissions. If you’d rather not upgrade, the GitHub app still works for notifications, issue triage, and PR management without the agent features.
The bottleneck problem#
Here’s the trade-off: if generating PRs becomes this easy, the bottleneck shifts from writing code to reviewing code.
That’s probably a good problem to have. Most teams have more backlog than capacity. If AI can tackle small fixes asynchronously, that frees developers for complex work.
But it only works if review capacity keeps up. If your team generates ten AI-authored PRs a day and reviews two, you just moved the backlog from issues to pull requests.
What this signals#
GitHub is building toward AI agents embedded everywhere you work: in your IDE, in GitHub, in Slack, in Teams (which also now supports this). The pattern is clear—they want Copilot available at the point where you realize work needs to happen, not just when you’re actively coding.
The bet is that lowering the friction to delegate work to AI will fundamentally change how teams operate. Instead of “we’ll fix that eventually,” it becomes “let’s have the agent fix that now and review it tomorrow.”
Whether that’s a productivity win or just creates more review load depends entirely on how good the agent is at getting things right on the first try.
Learn more: Check the GitHub Copilot Slack integration announcement for setup instructions. The GitHub app for Microsoft Teams now has the same capability.