Review Runners
Custom code review agents that run in GitHub Actions
Bring your model, your repo's review skills and scripts, and your team's reviewers. Version the review logic with your code and run it on every pull request. Pay StarSling for the runners; pay your provider for the model.
1 critical - 0 warning - 1 suggestion
Hosted review bots now let you customize their reviewer. Review Runners let you build your own.
Hosted code review bots support custom instructions, rules and repository context, but engineers still repeat themselves in every pull request. Review Runners put the review system itself in your repository: the model configuration, skills, scripts and reviewer responsibilities. Your team versions it with the code and runs it in CI.
Expensive
Your team may already pay for frontier or open-source models through an inference provider. A hosted review bot adds another AI bill. Bring the model account you already have and pay StarSling for the compute that runs the review.
Your review system
Your engineers already have skills, scripts and tools they use to review code. Commit that setup to the repository, define what each reviewer is responsible for, and run it on every pull request.
Bring your model, your review code and your reviewers. Run them in CI.
One workflow file. Your model, skills, scripts and team-defined reviewers run as GitHub Actions jobs on StarSling runners. Each reviewer leaves a review comment on the pull request, before a human opens it.
Your model, your bill
Bring your own provider key. The reviewer calls your model through your account. Your provider bills for the tokens; StarSling bills for the runner that executes the review job.
Your review code
Skills, scripts, instructions and reviewer configuration live in your repository, versioned with the code they review. A skill packages the context and scripts your team uses to review a part of the codebase. Merge an update and future reviews use it. Skills and instructions load from the trusted base branch, so a pull request cannot rewrite its own reviewer.
Your reviewers
Define reviewers for things like security, APIs, tests, database changes or frontend architecture. Give each its own responsibilities, skills and instructions. For example, you could give a security reviewer your security scripts and an API reviewer your API conventions.
Findings you can act on, one comment per pass
Each reviewer leaves one structured comment: a verdict, findings by severity, exact path:line references and a concrete fix. A counts line closes the comment, including zero counts when the reviewer finds nothing.
A new push cancels any pass still running on the previous commit, so the next completed review covers the latest code.
Pay StarSling for the runner. Pay your model provider for the model.
A review is a GitHub Actions job on a StarSling runner, billed by the minute at StarSling Runners pricing, in the same account as the rest of your CI. The model tokens are billed by your own provider, on the account you already have.
Runner minutes
A review runs on the 2 vCPU runner by default, at $0.004 a minute, the same rate card as every StarSling runner, from 2 to 64 vCPU. Billing runs from job start to finish, rounded up to the nearest minute; queue time is never billed, and concurrency is unlimited. Runner pricing.
Model tokens
Billed by your inference provider at your provider's rate, on the account you already have. The reviewer calls your model through your key.
Same platform, the other half of verification
Review the way your team reviews.
Tell us which repositories you want reviewed and which model you run, and we will set your org up.