Skip to content
NewsProduct launch

GitHub ships a Copilot code review API and locks in a costlier default effort level

· by Pondero Newsdesk

The short version

GitHub's October 2 changelog adds REST and GraphQL API support for Copilot code review and confirms that Balanced, which uses more AI credits than Lite, became the default effort level on September 28.

GitHub ships a Copilot code review API and locks in a costlier default effort level

Teams that run GitHub Copilot's automatic code review started spending more AI credits per review on September 28, when GitHub switched the default review effort from a lighter setting to "Balanced." Any repository or organization that never explicitly selected "Lite" is running Balanced now, and getting back to the lighter tier takes an explicit change at the organization or repository Copilot settings page, since repository settings override organization settings, which override enterprise settings. The October 2 changelog entry that documents the switch also ships something new: Copilot code review can now be requested directly through GitHub's own APIs instead of only from the pull request UI.

What happened

GitHub's October 2 changelog entry confirms two changes to Copilot code review, per GitHub's changelog. First, developers can now request a Copilot review through the REST and GraphQL APIs and set the review effort level on each request, so a review can start from a script, a CI pipeline, or an internal tool rather than only from the "Reviewers" box on a pull request. That API support is generally available on Copilot Pro, Pro+, Max, Business, and Enterprise plans. Second, "Balanced" is now the default review effort level for both new and existing repositories and organizations. The switch actually took effect September 28, 2026, following an announcement on August 28; any repository or organization that had explicitly set "Lite" kept that choice. Per GitHub's documentation on Copilot code review, Balanced routes a pull request to a higher-reasoning model for deeper analysis of complex logic, security-sensitive code, and cross-service changes, and uses more AI credits, and may consume marginally more GitHub Actions minutes, than Lite reviews.

Audit the four override levels before the bill shows up

Any repository or organization that left Copilot code review on its built-in default did not opt into Balanced. GitHub switched it for them on September 28, and the October 2 changelog is the first place that change is spelled out alongside the new API. GitHub's documentation splits the cost of a review into two components: AI credits for the model interaction itself, and GitHub Actions minutes for the agentic work the review does around it, such as gathering context from the rest of the repository. When a repository is configured to automatically request a review on every new pull request, that AI-credit consumption is attributed to the pull request's author rather than to the organization as a whole, so the switch to Balanced lands as a personal usage increase for whoever opens pull requests most often, not a single line item an admin can review centrally. That is exactly why admins should check the four override levels this week rather than wait for a usage report to surface the increase on its own: GitHub's configuration guide lists enterprise AI controls, organization Copilot settings, repository Copilot settings, and an individual's personal Copilot settings, with each level able to override the one above it. The new API carries the same choice forward. A review requested through the REST or GraphQL endpoint can set its own effort level per call, so a CI job that calls the API without specifying "Lite" lands on Balanced by default, the same as a review requested through the pull request UI.

Context

Automatic review itself is a separate setting from the effort level, and GitHub's documentation describes three ways to turn it on: a user can configure Copilot to automatically review pull requests they personally open, a repository owner can configure Copilot to review every pull request in that repository, or an organization owner can configure Copilot to review pull requests across some or all repositories in the organization. Automatic review is available on the Copilot Pro, Pro+, and Max plans, and with a Business or Enterprise license, per GitHub's documentation. Copilot code review also carries agentic capabilities that run on top of whichever effort level is selected: full-project context gathering, which analyzes the rest of a repository to inform a review rather than just the diff, and, in public preview, the ability to pass a review's suggestions directly to Copilot's cloud agent, which then opens a new pull request with the fixes already applied. Those capabilities are enabled automatically on every plan that includes code review and are not affected by the Lite-versus-Balanced choice; they run through the Actions-minutes side of the cost split rather than the AI-credit side.

What to watch next

A third effort level, "Max," is still listed as "Coming soon" and not yet available, per GitHub's documentation. The actionable step is not to wait on that rollout, or on GitHub reconsidering the default: a team that wants Lite's lighter review back has to set it now, at the organization or repository Copilot settings page. Repository settings override organization settings, which override enterprise settings, so the fastest fix for a single repository is that repository's own Copilot code review setting.

Sources