Runrailby Veyact
Try Runrail

Workflow

From candidate branch to review packet.

Runrail materializes the candidate in an isolated local review directory, enforces the base-owned review boundary, and returns six primary artifacts plus focused review and hunk-attention evidence. It stops before source apply.

  1. 01

    Choose the review

    Select an existing AI candidate branch and its clean base commit.

  2. 02

    Bind the review contract

    Runrail reads policy only from the selected base commit, records required evidence, and blocks candidate attempts to weaken it.

  3. 03

    Check locally

    The review happens in an isolated local copy using non-executing Python, JavaScript, JSON, TypeScript, TSX, and JSX syntax or parse checks. Unsupported changed source types block the preflight.

  4. 04

    Track attention and decide

    Record exact-hunk attention, compare a related revised audit when available, then read the patch and evidence packet. Apply remains a separate human-controlled action.

Runrail Recorded local assurance loop
No sign-upLocal evidence only
Assurance loop

Review one AI-generated change

Recorded product tour
Repositorypayments-service
Basemain
Candidateai/refactor-checkout
Derived verdict

Needs attention

Checks completed. Three items still need a human decision.

  1. 01
    Inspect the security-sensitive changeOne changed area matches the base-owned review contract.
    1 item
  2. 02
    Resolve unreviewed hunksChanged or ambiguous hunks never inherit prior attention.
    2 hunks
  3. 03
    Record approved project testsRunrail accepts the evidence; it does not execute the suite.
    Not run
Inspect evidenceCopy agent briefRerun and verify
No upload.No automatic apply.Human decision required.

Technical guide

See commands, outputs, and limitations.

Open Docs