GitHub Copilot enters Xcode’s built-in assistant in public preview

A new integration, alongside the existing extension

GitHub announced on November 4 that Copilot can now be used through Xcode’s built-in coding assistant. The integration is in public preview beginning with Xcode 27.3 beta. GitHub says it supports all Copilot plans and requires a Mac running macOS Tahoe 26.6 or later.

Setup is through Xcode’s Settings, under Intelligence. The existing Copilot for Xcode extension remains a separate option. This is therefore an additional way to work with Copilot, not an announcement that the extension has been retired. Read The Day checked GitHub’s release note but has not installed or tested the beta. Source: GitHub Copilot is now available in Xcode’s coding assistant

Why the location of an assistant matters

Our analysis: bringing an assistant into the development environment can reduce the handoff between asking for help and inspecting the resulting change. A developer can evaluate the proposed code in the context of the project rather than manually moving fragments between an editor and a separate conversation.

That convenience should be measured against the workflow already in use. If the current extension meets a team’s needs, a new integration is not automatically an improvement. The useful comparison is whether it makes a representative task easier to complete and review without introducing new failure modes.

A good trial might involve a small bug with a clear reproduction, a limited refactor and a request to explain unfamiliar code. Those tasks exercise different abilities. Success at generating a new screen does not establish reliable behavior when modifying an existing application.

Keep preview evaluation separate from release work

The beta requirement is material. Teams should avoid treating a preview integration as a reason to change a production release toolchain without evaluating the consequences. A separate test project or controlled branch can reveal whether the assistant fits the workflow while keeping the trial bounded.

Our recommendation is to record the exact Xcode version used in the trial. If the assistant proposes code that depends on a newer SDK, the reviewer needs to distinguish an AI integration issue from a platform compatibility issue. Reproducible environment details make feedback more actionable.

The announcement does not establish feature parity between the built-in assistant and the separate extension. Teams should compare the actions they actually rely on instead of assuming that shared branding means identical behavior, settings or controls.

Review the change, not just the conversation

An assistant’s explanation can sound complete even when the patch does not address the original problem. The acceptance decision should therefore return to the application’s behavior. Does the issue reproduce before the change? Does the intended behavior work afterward? Did the patch alter another path unexpectedly?

For interface work, a successful build is only one check. Layout, accessibility and state transitions may need inspection in the running app. For data-handling code, the important checks may concern error paths and persistence. The right validation depends on the task rather than on which assistant generated it.

Reviewers should also be able to identify all modified files. Keeping the request narrow makes it easier to spot unrelated cleanup that increases the scope of the change without helping solve the problem.

Adoption should follow a useful comparison

A team evaluating the preview could track total completion time, the amount of manual correction and the quality of the final explanation. It should include unsuccessful attempts instead of selecting only the most impressive session.

Account and organizational settings also deserve an explicit check before sensitive project material is used. Support for a subscription plan does not establish that a particular employer has enabled the integration or approved every associated workflow.

The release is useful for developers who want Copilot inside Xcode’s native assistant experience. The sensible next step is a small, repeatable evaluation against the current extension. The result should determine whether to adopt the new surface, while preview status remains clearly visible in that decision.