GitHub Copilot Coding Agent: Ask @copilot to Modify Any Pull Request
GitHub has expanded the Copilot coding agent to allow developers to request code changes on any pull request — including those authored by humans — by simply mentioning @copilot in a comment. Previously, Copilot would always respond by opening a new pull request on top of the existing one; now it makes changes directly to the same branch. The agent runs in its cloud-based environment, validating changes with tests and linters before pushing, and is available to all paid Copilot plan subscribers.
Sources & Mentions
2 external resources covering this update
Ask @copilot to Make Changes to Any Pull Request
GitHub has made a meaningful improvement to how the Copilot coding agent interacts with pull requests. Developers can now mention @copilot in a comment on any pull request — whether originally authored by a human or by Copilot itself — to request targeted code modifications. This removes a significant friction point in the previous workflow and brings Copilot directly into the iteration loop of everyday development.
What Changed
Before this update, when @copilot was mentioned in a pull request comment, Copilot would respond by creating a new, separate pull request on top of the existing one. That behavior sometimes created unnecessary branching overhead, particularly when making small iterative changes like fixing a failing test, addressing a reviewer's comment, or adding a missing unit test.
With this release, Copilot now edits the existing pull request's branch directly. The result is a tighter feedback loop: developers can leave a natural-language instruction — such as "Fix the failing GitHub Actions workflow" or "Add a unit test covering the case when the model argument is missing" — and Copilot will work within the existing branch to apply that change.
For teams that prefer the previous behavior, the option remains. Asking Copilot with explicit language — for example, "Open a PR to fix the failing tests" — still triggers the child pull request workflow.
How the Agent Works
Copilot's coding agent operates in an isolated, cloud-based development environment. When triggered via @copilot, it clones the repository, analyzes the context of the pull request, makes the requested code change, and validates it against the repository's tests and linter before pushing back to the branch. This validation step is built-in and requires no configuration from the developer.
Access is limited to users with write access to the repository. Mention-based triggering is restricted to this group specifically, ensuring that only authorized contributors can direct the agent's activity.
Availability
The feature is available to all paid Copilot plan subscribers: Pro, Pro+, Business, and Enterprise. Copilot Business and Enterprise users require an administrator to have enabled the Copilot coding agent feature in organization settings before it becomes accessible.