Agent Upstart

Cool things people are building with agents — and practical help getting the most out of them.

Playbooks · Coding tasks

Copilot code review goes Adaptive, checks secrets

Written November 18, 2026

Copilot code review now picks Lite, Balanced or Max effort per pull request with Adaptive, draws on related repositories, and flags exposed secrets.

Until now, Copilot code review made you choose one review effort and live with it for every pull request: Lite for speed, Max for depth, the same setting for a documentation edit and an authentication change alike. GitHub's changelog post of November 17, 2026 replaces that fixed choice with Adaptive, which picks the effort per pull request. The same release lets reviews draw on related repositories, adds secret detection inside reviews, and brings Copilot code review to Azure DevOps as generally available. Adaptive, cross-repository context and secret detection are in public preview, and features in public preview are subject to change. All three are available to Copilot Pro, Pro+, Max, Business and Enterprise plans.

Adaptive picks Lite, Balanced or Max for you

Adaptive matches review depth to each pull request's complexity and risk, then reviews it with Lite, Balanced or Max. A documentation update may only need Lite, while a change to authentication logic benefits from Max, and Adaptive makes that call instead of leaving it to a setting someone chose months ago. Each review shows which effort level Adaptive picked, so the choice stays visible rather than hidden inside the review. Balanced remains the default review effort level for anyone who does not turn Adaptive on. One guardrail matters for administrators: Adaptive never picks a level above the maximum review effort level set for your enterprise, organisation or repository, so the ceiling you configured still binds.

Where Adaptive is set

You can select Adaptive for a single pull request from the Reviewers bar, which suits a one-off risky change on a repository that normally runs light. Or you can set it as the default review effort in personal, repository, organisation or enterprise settings, which suits a team that wants the per-request judgement everywhere without asking reviewers to remember it. The two routes stack with the ceiling above: the maximum review effort level configured in Copilot code review settings determines which levels Adaptive can choose from. If your organisation capped reviews at Balanced, Adaptive chooses between Lite and Balanced only, no matter how risky a pull request looks.

Reviews can read related repositories

Changes often affect code outside the repository you are working in, and a review that can only see one repository misses that. With cross-repository context, Copilot code review can read related public repositories during a review, so it can find code in other repositories that depends on your change. It can also find relevant context from documentation stored in other repositories, custom instructions and skills, and more. Cross-repository context is in public preview for public repositories. For a team whose libraries and services live in separate repositories, this is the difference between a review of a diff and a review of what the diff touches.

Secrets get flagged inside the review

Copilot code review can now flag exposed secrets in your pull requests, pointing out exposed credentials in review comments so you can remove them before you merge, rather than after they reach the main branch. It works two ways. Known formats are covered through GitHub's secret scanning partnership programme: more than 150 technical partners contribute detectors for over 500 secret types, using deterministic methods such as regular expressions. Secrets without a recognizable format are covered the other way, with AI and machine learning detection that uses large language models and GitHub's purpose-built model to find exposed secrets. To turn secret detection on, go to your repository or organisation settings and select Copilot, then Code review.

Azure DevOps, and one line on outcomes

Copilot code review for Azure Repos, previously in public preview, is now generally available, so teams hosting code in Azure DevOps can get Copilot reviews on their pull requests. Administrators turn it on at the organisation, project and repository level, and can add a branch policy to request a review automatically, or reviewers can request one manually from the Reviewers list. A project or repository default is Lite or Balanced, custom instructions and skills tailor the feedback, and usage is billed in GitHub AI credits to the Azure subscription linked to the Azure DevOps organisation. Separately, review outcomes can now be tracked per repository in Repository insights, covered in its own companion post the same day.

What to do

None of this changes what a review costs you in attention when Adaptive guesses wrong, and the preview label means the behaviour can shift. But the fixed-effort compromise was the weakest part of the tool, and it is the part this release removes.

Related playbooks

Sources

Back to: Coding playbooks hub · Also: Decide if Copilot Code Review Max is worth it