GitHub shipped four code review upgrades on November 17 under the Microsoft Ignite banner. The headline feature is Adaptive Effort — Copilot now decides how deeply to review each pull request on its own. Alongside that, it added secret detection before merge, pushed Azure DevOps code review to general availability, and introduced cross-repository context for public repos. Here is what actually changed and what you need to pay attention to.
Adaptive Effort: Copilot Calibrates Its Own Review Depth
Until now, teams set a fixed effort level — Lite, Balanced, or Max — and Copilot applied it uniformly across every pull request. With Adaptive, Copilot evaluates each PR’s complexity and risk and picks the appropriate level automatically. A trivial config change gets Lite. A PR touching authentication logic gets Max. The goal is obvious: stop burning AI credits on boilerplate changes and spend them on code that actually needs scrutiny.
Adaptive is currently in public preview and opt-in. You can set it as your default in personal, repository, organization, or enterprise settings, or choose it per PR from the Reviewers bar. One important constraint: Adaptive respects whatever ceiling your organization has configured. If your enterprise caps effort at Balanced, Adaptive cannot go higher than that regardless of how risky it thinks a PR is.
The feature works, but GitHub has not disclosed how Copilot scores complexity and risk. That matters. Teams running Adaptive should monitor which effort level it assigns and verify it is behaving sensibly — especially for changes in security-sensitive code. Each review does show which level was chosen, so the data is there. Use the new Repository Insights dashboard to track whether Copilot’s reviews are actually being actioned.
Secret Detection: Catch Credentials Before They Merge
Copilot code review can now flag exposed secrets directly in pull request comments. The detection runs before merge — which is a meaningful distinction from GitHub’s existing secret scanning, which triggers after code is committed to the repository.
The system uses two layers. First, a deterministic regex-based scanner covering 500+ known secret formats — built through GitHub’s partnership program with over 150 technical partners. This catches AWS keys, GitHub tokens, Stripe and Twilio credentials, and hundreds of others with high confidence. Second, an ML model that handles secrets without a recognizable pattern. That second layer is what makes this more than just repackaged secret scanning.
The practical value: developers working in test fixtures, config files, or example code occasionally slip real credentials in. Surface that in a review comment and there is a natural moment to fix it before it reaches the default branch. Secret detection is in public preview and available on all Copilot plans. See the secret scanning on-demand changelog for the full scope of what is covered.
Azure DevOps Code Review Is Now Generally Available
Copilot code review for Azure Repos has cleared general availability after spending two months in public preview. For teams on Azure DevOps, that means a supported SLA, custom instructions via .azuredevops/copilot-instructions.md, and billing through GitHub AI credits charged to your linked Azure subscription.
Costs run $0.05–$1 per review at Lite effort and $0.25–$5 at Balanced. Charges appear in the Azure portal with roughly a 48-hour delay. For a team merging 50 pull requests a week at Balanced, budget up to $250 per week at the high end. Set a budget alert in Azure Cost Management before you roll this out at scale.
Cross-Repository Context and What Comes Next
Also in public preview: Copilot can now read related public repositories during a review to find downstream code that depends on your change. Library authors and monorepo teams managing shared packages will find this useful — Copilot can tell you if your change breaks an interface something else relies on. You choose which repositories Copilot can access.
The bigger picture here is that Copilot is steadily moving from “suggests code” to “reviews code” to “decides how to review code.” Adaptive Effort is an early version of autonomous PR triage. The next logical step is Copilot flagging which PRs need human review at all — that is not today, but the trajectory is clear. For now, enable Adaptive on a non-critical repository, monitor the effort distribution, and treat secret detection as an extra safety net rather than a replacement for a proper secrets manager.
All four features — Adaptive Effort, Secret Detection, Azure DevOps GA, and Cross-Repository Context — are available on Copilot Pro, Pro+, Max, Business, and Enterprise plans. Full configuration details are in the GitHub Docs and the official changelog entry.

