Runrail Desktop 2.4.14
Checking local runtime

Local activation

Prepare this device.

Confirm the local runtime and entitlement before reviewing a repository.

Entitlement metadata only No repository upload
Runtime Checking

Integrated doctor is running.

Entitlement Checking

Local access state is loading.

Boundary Local and human-controlled

No source apply, arbitrary shell, or background scanner.

First review checklist

Three steps to local evidence.

Checking this device.

  1. 01
    Confirm this deviceIntegrated doctor checks the local package.
    Current
  2. 02
    Complete the safe sampleProve the real loop on a disposable fixture.
    Pending
  3. 03
    Open local evidenceSelect a repository or resume verified history.
    Pending

Guided sample

Review a safe local fixture

No enrollment or repository path required.

Daily assurance

Today

Select a local repository or run the safe sample.

Current assurance loop

Complete one evidence loop.

0 completed this week 0 active repositories
  1. 01
    ReviewInspect candidate evidence
    Current
  2. 02
    InspectRecord hunk attention
    Pending
  3. 03
    ResolveVerify any correction
    Pending
  4. 04
    DecideReview the human decision
    Pending

0 of 4 stages complete. Select a repository or run the safe sample.

Repository
None selected
Last verdict
No review
Unresolved attention
0
Assurance phase
Not started
Expiring exceptions
0
Review Inbox 0 open

No unfinished local reviews.

Review a candidate change

Compare two local Git references, inspect deterministic evidence, and leave the final decision to a human.

No upload No automatic apply

References remain editable. Inspecting only reads repository metadata.

Ready. Source apply is unavailable.

Awaiting review

Run a review to populate this decision surface.

Evidence, remaining actions, and local artifacts will appear here.

Next actions Up to three
  1. 1
    Select a repository and candidate

    Runrail will derive a bounded local review plan.

No review selected.

Agent Assurance is ready after the first review.

Advanced continuity

Continuity and provenance

Optional local tools remain separate from the core review decision.

Explicit actions only Authority none
Review FocusDeterministic order for the work that remains 0
  1. Review Focus unavailable

    Run a candidate review to create an ordered plan.

FindingsDeterministic issues and human dispositions 0
  1. No finding recorded

    Project tests and human review are still required.

Changed filesCandidate paths reported by the local audit 0
  • No changed file list yet.
Hunk attentionExact local attention state, never approval 0

No hunk coverage yet.

  1. No coverage loadedRun a review to inspect exact diff hunks.

Run a review to create path-free coverage.

Revision Pro feature / Desktop Revision Workspace Not mapped
Current session audit Run a current review first.

Runrail automatically selects the latest related local session after you confirm a rerun. No repository is uploaded and no project test is executed.

No map yet

Carry exact attention, reset everything else.

The result will show relationship, policy continuity, finding changes, and current attention required.

Relationship
-
Policy basis
-
Carried exactly
0
Current attention
0
Added findings
0
Resolved findings
0
    Re-review Map artifact No Re-review Map path yet.
    Agent AssuranceGive exact objectives to any coding agent, then verify the next revision Not generated

    The handoff contains fixed, source-free correction objectives. A fixed local adapter can open only after confirmation; Runrail never supplies agent arguments, trusts completion, changes policy, applies a patch, or approves a merge.

    Copy remains available for every coding agent. Interactive launch requires a detected fixed adapter and terminal.

    Local packet No Agent Assurance packet yet.
    Local result Rerun after a revision to create the result.

    No result yet

    Close the loop with evidence, not agent promises.

    After a rerun, Runrail will classify resolved, remaining, new, and reintroduced findings. Ready for human decision also requires complete Review Coverage.

    Resolved
    0
    Remaining
    0
    New
    0
    Reintroduced
    0
      External evidence receiptsImport aggregate SARIF, JUnit, or LCOV evidence without executing it 0

      Choose evidence created for this exact candidate revision. Runrail verifies the checksum and revision, strips paths and raw output from the receipt, and never lets imported evidence change the verdict by itself.

      Run a review before importing external evidence.

      No external evidence receipts are attached to this review.

      Agent Outcome LedgerRepository-local verified rerun outcomes with sample uncertainty Pro

      Record which agent worked on this verified rerun. Runrail stores source-free outcome counts only and never creates a global model score or claims one agent writes better code.

      Agent Outcome Ledger requires Pro and a verified rerun.

      Samples
      0
      Agent labels
      0
      Ready outcomes
      0
      Evidence complete
      -

      No verified agent outcomes are recorded for this repository.

      Local product metricsTransparent weekly assurance-loop counts stored only on this device Local

      Runrail measures completed assurance loops so repeated reviews get easier to understand. Raw events remain local. No repository identity, path, source, diff, prompt, finding, agent output, or provider data is uploaded.

      Loading this week’s local assurance activity...

      Completed loops
      0
      Active repositories
      0
      Loops per repository
      0
      Upload
      Off

      Optional daily aggregate upload is implemented server-side but remains disabled by default. This Desktop release sends nothing unless a future explicit-consent control is separately enabled; repository identity and source-bearing data are never accepted.

      Candidate TournamentDeterministic review order across two to five local branches Pro

      Choose local branches already discovered ahead of the same confirmed base. Runrail ranks which candidate deserves human review first; it never selects the best code.

      Local candidate branches

      Inspect a clean repository to discover tournament candidates.

      Requires Pro and two to five discovered local branches.

      No tournament yet

      Compare review attention, not code quality.

      Every candidate uses the same base-owned contract and safe-static verifier. The final choice remains human.

      Why review this candidate first Run a Tournament to create an equal-basis review-order receipt.
        Path-free tournament artifact No Candidate Tournament path yet.
        HistoryPrior local reviews for this repository 0

        No local review history is available.

        Repository Trust ProfileLocal review continuity and weekly engineering evidence Pro

        Trust Profiles are derived locally from verified review history. They never rewrite the base-owned review contract.

        Review cadence
        Not loaded
        Current contract
        Not loaded
        Verdict changes
        0
        Attention closed
        0
        Expired exceptions
        0
        Passport continuity
        0 / 0

        Select a reviewed repository to load local continuity.

        ProofGraphFindings, rules, evidence, hunks, dispositions, and revisions Pro

        Build a path-free local graph from the verified review lineage. Explicit dispositions classify findings but never hide them or rewrite policy.

        Revisions
        0
        Findings
        0
        Hunks
        0
        Rules
        0
        Dispositions
        0
        Hidden
        0
        Verified relationships Path-free counts
        1. No graph loadedRun a review first.
        Path-free local ProofGraph No ProofGraph built yet.

        Requires Pro and a verified local review session.

        AdvisorOptional secondary opinion with authority none Optional
        Advisory provider

        Run a review first. Authority remains none.

        No advisory yet

        Fixed review candidates or a safe abstention.

        The advisor cannot create commands, patches, apply actions, or merge approval.

        Local boundary No upload. No automatic apply. Human decision required.