Copilot code review picks its own effort level per PR
GitHub's November 17, 2026 release adds Adaptive review effort, secret detection in pull requests, cross-repository context, and general availability for Azure Repos.
GitHub lets the reviewer size up the change
GitHub shipped a Copilot code review update on November 17, 2026 that lets the tool decide how hard to look at each pull request. Adaptive sits alongside secret detection and cross-repository context, and Copilot code review for Azure Repos moves from preview to general availability in the same release. According to the GitHub Changelog announcement, the new capabilities are in public preview and available to Copilot Pro, Pro+, Max, Business and Enterprise plans; GitHub notes that preview features are subject to change.
Adaptive chooses Lite, Balanced or Max
Before this release, a person chose the effort level in settings or when requesting a review. GitHub's example: a documentation update may only need Lite, while a change to authentication logic benefits from Max. Adaptive reads each pull request's complexity and risk, then reviews at Lite, Balanced or Max, and each review shows the level it chose. Adaptive can be selected for one pull request from the Reviewers bar, or set as the default in personal, repository, organization or enterprise settings. Balanced remains the default, and Adaptive never exceeds the maximum effort level set for the enterprise, organization or repository.
Secrets, related repositories, outcome tracking
Secret detection uses two methods. Deterministic detectors cover known formats: GitHub says it works with more than 150 technical partners through its secret scanning partnership program, and the resulting 500+ secret types are matched with techniques such as regex. A second layer uses LLMs and a purpose-built GitHub model to catch credentials with no recognizable format. Findings appear as review comments so a key can be pulled before merge. It is switched on in repository or organization settings under Copilot, then Code review.
Cross-repository context lets a review read related public repositories to find code elsewhere that depends on the change, plus documentation, custom instructions and skills stored in other repositories. It is in public preview for public repositories.
The Insights tab of a repository, under Repository insights, now shows how Copilot review feedback is addressed and used.
Azure DevOps moves to general availability
Copilot code review for Azure Repos is now generally available after its public preview. Administrators enable it at organization, project or repository level; a branch policy can request reviews automatically, or a reviewer can ask from the Reviewers list. Projects and repositories can default to Lite or Balanced, with per-review overrides, and custom instructions and agent skills shape the feedback. Applying and committing a suggestion resolves the related comments in the same step, and audit log events record when the feature or its settings change. Usage is billed in GitHub AI credits to the Azure subscription linked to the Azure DevOps organization, so costs appear in Azure Cost Management. The source lists Lite and Balanced here and does not mention Adaptive, Max, cross-repository context or secret detection for Azure DevOps.
What this changes for people building apps
If much of your code is written by an agent or assistant, the effort decision is the one you are least likely to make deliberately. Sending a README fix and a login change through the same setting is either wasteful or shallow, so the likely benefit is that a risky pull request gets the deeper pass without anyone remembering to request it. Secret detection addresses a related problem: code generated quickly is committed quickly, and a credential pasted into a config file is cheaper to catch at review than after deploy. That reading is inference. The source describes capabilities, not measured results.
Limits and open questions
Three of the items are public previews, and GitHub says preview features are subject to change without giving dates for general availability. The source does not say whether cross-repository context will reach private repositories, how often Adaptive lands on each level in practice, or how secret detection inside reviews relates to the secret scanning alerts teams already receive. Pricing for the preview features on GitHub plans is not stated, though the Azure DevOps billing model is described.
For AI app builders
Apps produced with an AI app builder often arrive as a few large, fast changes, which is the shape of pull request a fixed review setting handles worst. Draftship's paid plans include GitHub push, so a project's code can live in a repository where review tooling of this kind runs; the GitHub and Vercel docs cover that path. Nothing in the announcement changes how Draftship builds or hosts an app, and Draftship is not affiliated with GitHub.
Primary sources
- Copilot code review: Adaptive effort, secret detection, and more · GitHub Changelog, Nov 17, 2026
This article was drafted with AI assistance from the primary sources listed below and checked automatically against them. An editor hasn't reviewed it yet.