Runrailby Veyact
Try Runrail

Independent assurance for AI-generated code

Runrail

Review AI code with evidence, not trust.

Runrail reviews an existing candidate branch on your machine, turns findings into a brief for any coding agent, and verifies the next revision before you decide what ships.

Try the sample review Download Runrail preview

No sign-up for the recorded review. The preview includes doctor, the safe sample, a path-free Repository Readiness Receipt, and a Review-Ready Brief you can copy or download.

Recorded product tour Review → Brief → Verify No sign-up
Runrail Recorded local review

Repository stays on this device

Candidate review

Checkout refactor

Local evidence · 10:42
Derived verdict

Needs attention

Three items need a human decision before this change is ready.

  1. 01
    Inspect the security-sensitive hunkMatched by the base-owned review contract.
    1 item
  2. 02
    Resolve unreviewed changesChanged hunks never inherit prior approval.
    2 hunks
  3. 03
    Record project-test evidenceRunrail records evidence; it does not run the suite.
    Not run
NextInspect the first attention item
One evidence layer for CodexClaudeGeminiGrokany Git branch
  • Your code stays localNo repository upload
  • Every verdict shows its workChecksummed evidence
  • You own the final decisionNo automatic merge

One repeatable habit

Review. Hand back a brief. Verify the revision.

Runrail gives every AI coding session the same finish line: visible evidence and a human decision.

  1. 01

    Review the candidate

    Select the base and AI branch. Runrail checks supported files locally and shows the three actions that matter most.

    Repository stays local
  2. 02

    Hand off exact objectives

    Copy a source-free correction brief to Codex, Claude, Gemini, Grok, or the agent you already use.

    Agent authority: none
  3. 03

    Verify, then decide

    Rerun the revision and see what resolved, remained, appeared, or returned before making the human decision.

    Evidence over claims

Why you come back

The evidence gets more useful with every rerun.

Runrail remembers the proof chain, not your source code. Reopen the exact review, see what changed across revisions, and measure outcomes without trusting an agent's completion message.

  • Exact historyRehydrate the verified audit, coverage, and assurance state.
  • Agent-neutralUse the same finish line across every coding agent.
  • Portable proofKeep checksum-linked receipts without exporting repository identity.
  1. Review 013 attention itemsCandidate evidence captured
  2. CorrectionBrief sent to your agentNo source exported
  3. Review 023 resolved · 0 newRevision verified
  4. DecisionHuman review readyMerge authority unchanged

Start simple. Add continuity when it matters.

A complete individual loop, with a clear path to deeper evidence.

No metered local reviews. No mandatory model. Your evidence remains portable.

Starter · EUR 19.99/month

Close every review loop.

One device, unlimited local reviews, exact history, Review Coverage, Agent Assurance, external evidence receipts, and checksum Change Passports.

Pro · EUR 49/month

Build a local evidence memory.

Three devices, Trust Profiles, weekly reports, signed passports, agent outcome history, Candidate Tournament, and ProofGraph.

Compare Starter and ProBoth are implemented and delivery-ready. Purchase remains independently paused.

Your stack. Your agent. Your machine.

Try the complete loop before you choose a plan.

One independent review layer across every AI coding workflow. Use Runrail with Codex, Claude, Gemini, Grok, or another agent. The public preview turns read-only repository inspection into one deterministic readiness state and next action; real-repository review requires a paid entitlement after trusted installers and commerce checks are complete.

Try the sample review Download Runrail preview
  • WindowsDesktop preview
  • macOSDesktop preview
  • LinuxDesktop and CLI preview
  • WSL2CLI preview