Feature · Work

PR trials

Compare past pull requests by result, checks, time and estimated cost.

Last changed

Generated from commit be57775 ·

How it works

What a trial does

  1. You pick a merged GitHub.com pull request and give the original requirement.
  2. The runner on your computer rebuilds the repository's history up to before the change. The model does not see the merged change.
  3. Your own Claude Code or Codex builds a candidate within the turn, time and cost caps you set.
  4. When the candidate is frozen, the checks run without a model on separate copies of the base, the candidate and the reference.
  5. You compare the two diffs and record a verdict: Candidate preferred, Reference preferred, Comparable or Inconclusive.

What you need

  • Agent runner 0.5 or newer on macOS, with a working sandbox;
  • a merged pull request on GitHub.com;
  • checks that run offline with the tools you already have.

Statuses

Ready to run, Queued, Running, Verifying, Completed, Failed and Cancelled.

Cost

The trial uses your existing CLI login on your computer. Agent does not give model credits for it. Costs shown are estimates, not bills.

From the help article PR trials, checked against the product on .

What you can do

The operations this feature's screens call. Each one is POST /api/<name>, and the badge says who may call it.

  • prTrials/prepare

    MemberNo API tokens

    PR trials: prepare.

    Types: PreparePrTrialRequest returns { trial: PrTrialDto }

  • prTrials/start

    MemberNo API tokens

    PR trials: start.

    Types: StartPrTrialRequest returns { trial: PrTrialDto }

  • getRunnerTask

    Member

    Gets the runner task.

    Types: GetRunnerTaskRequest returns RunnerTaskDetailDto

  • prTrials/review

    MemberNo API tokens

    PR trials: review.

    Types: { id: string; verdict: PrTrialVerdict; rationale: string; humanMinutes: number } returns { trial: PrTrialDto }

  • prTrials/cancel

    MemberNo API tokens

    PR trials: cancel.

    Types: { id: string } returns { trial: PrTrialDto }

  • prTrials/export

    Member

    PR trials: export.

    Types: { id: string } returns { report: PrTrialDto; protocolVersion: string }

  • prTrials/readiness

    Member

    PR trials: readiness.

    Types: Record<string, never> returns PrTrialReadiness

  • prTrials/list

    Member

    PR trials: list.

    Types: Record<string, never> returns { trials: PrTrialDto[] }

  • prTrials/delete

    MemberNo API tokens

    PR trials: delete.

    Types: { id: string } returns { deleted: true }

  • prTrials/get

    Member

    PR trials: get.

    Types: { id: string } returns { trial: PrTrialDto }

Screens

What it records

The feature writes these events to the audit log, the permanent record of what happened.

  • Integration changed

Generated from code at commit be57775 ·

Turn the numbers into shipped work.

Agent runs these choices for you: a persistent AI worker with memory and rules, on your Claude and Codex subscriptions.