Review Runners vs GitHub Copilot code review

On a pull request, open Reviewers in the right sidebar and click Request next to Copilot (Using GitHub Copilot code review). If your organization provides Copilot, it enables the Copilot code review option in its Copilot policy settings (About GitHub Copilot code review).

For automatic review, turn on Automatic Copilot code review in your Copilot settings for the pull requests you create, or add a branch ruleset with Automatically request Copilot code review to cover a repository (Configuring code review by GitHub Copilot).

Copilot reviews the pull request, identifies issues and suggests fixes you can apply in a couple of clicks (About GitHub Copilot code review). It posts as copilot-pull-request-reviewer[bot], labels each comment High, Medium or Low, and leaves a Comment review by default (Using GitHub Copilot code review).

Review effort comes in two levels: Lite, a standard review, and Balanced, the default, which routes the pull request to a higher-reasoning model. Copilot can also use the repository's agent skills and MCP servers when they are relevant to the review (About GitHub Copilot code review).

Copilot code review runs its agentic capabilities, including full project context gathering, on GitHub Actions. It uses standard GitHub-hosted runners by default, and can use larger GitHub-hosted runners or self-hosted runners on Actions Runner Controller, on Ubuntu x64 Linux (About GitHub Copilot code review, Configuring runners).

A Review Runner is a GitHub Actions job in the workflow you already have, on the starsling-review runner. The job reads the repository with contents: read and holds pull-requests: write to leave its review as comments on the pull request.

Copilot code review runs on a mix of models, prompts and system behaviors that GitHub tunes for review, and GitHub selects the model for each review automatically. AI credits are charged to the person who requests the review, or to the pull request author when a policy triggers it (About GitHub Copilot code review, Models and pricing).

A Review Runner calls the model you choose through your own provider key, on your own provider account. AI code review in GitHub Actions walks through model choice and reviewer context.

Hosted review bots let you customize their reviewer. Review Runners let you run your own.

Copilot code review reads repository-wide rules from .github/copilot-instructions.md, path-specific rules from *.instructions.md files under .github/instructions/, shared rules from AGENTS.md, and the repository's agent skills, and organization owners can add custom instructions in organization settings (About GitHub Copilot code review, Adding repository custom instructions). Copilot reads them from the pull request's head branch, so a change to them is tested in the same pull request.

With Review Runners, the reviewer is yours: its skills and instructions live in your repository and load from the trusted base branch, so a change to a reviewer takes effect once it merges.

Each Copilot code review is billed in two ways: tokens in AI credits, and GitHub Actions minutes for the agentic infrastructure, with the minutes attributed to the repository (Models and pricing).

Review Runners bill per minute for the starsling-review runner that executes the review job, and your model provider bills your own key for the tokens. How to run automated code review in GitHub Actions covers the setup and the cost of a review job.