No-sign-up product walkthrough
Try Runrail with the tools you already use.
See one candidate move from attention to a source-free correction brief, a verified rerun, and a human decision. The walkthrough is recorded; no account, repository, upload, or model is required.
- No sign-up
- Any coding agent
- No repository upload
- Human decision
Optional setup
Use the machine and coding agent you already have.
These two choices personalize only this recorded walkthrough. They stay in this tab and are never stored, uploaded, or sent to a model provider.
See the same Review, Brief, and Verify loop with the coding agent you already use.
Review → Brief → Verify
See the decision change with the evidence.
Use one next action to move through Review, Brief, and Verify. The recorded evidence changes with each step, but no code runs and no network request leaves this page.
Review one AI-generated change
Needs attention
Checks completed. Three items still need a human decision.
- 01Inspect the security-sensitive changeOne changed area matches the base-owned review contract.1 item
- 02Resolve unreviewed hunksChanged or ambiguous hunks never inherit prior attention.2 hunks
- 03Record approved project testsRunrail accepts the evidence; it does not execute the suite.Not run
Correction brief ready
Give fixed objectives to any coding agent without exporting source, paths, or repository identity.
- 01Preserve the base-owned security ruleRevise the candidate without changing the review contract.Required
- 02Resolve both attention-bound hunksThe next review must prove each changed hunk again.2 hunks
- 03Return external test evidenceThe agent may suggest work; Runrail trusts only the rerun evidence.External
Ready for human decision
The revision closed the recorded objectives. Runrail still does not approve a merge.
- Resolved
- 3
- Remaining
- 0
- New
- 0
- Reintroduced
- 0
What brings you back
Return to the next decision, not another dashboard.
Runrail keeps the verified local review state useful across revisions. It does not manufacture streaks, scan in the background, or upload product telemetry.
- 01Resume the exact review
Today restores the current verdict, unresolved attention, Assurance phase, and one primary action.
- 02Work only the open queue
The local Review Inbox shows unfinished reviews, returned findings, and expiring exceptions. Nothing scans in the background.
- 03Prove what changed
Related reruns show resolved, remaining, new, and reintroduced findings while completed-loop metrics stay on your machine.
No dark patterns. Repeat use is earned by a shorter, verifiable path from an AI-generated change to a human decision.
Advanced evidence inspector
Open a recorded or local audit.json in this browser
Browser memory only · no upload · no backend · no execution
Content checksums detect accidental changes. They do not authenticate who created either file.
Start with recorded evidence or open your own audit.
The bundled sample is verified in this browser. Local files can also be chosen or dropped here. Maximum size: 1 MB.
Audit could not be verified.
The file is invalid, unsupported, or its checksum does not match.
- Content checksum
- Matches
- Preflight
- Passed
- Static checks
- Passed
- Baseline contract
- Passed
- Project tests
- Not run
- Source checkout
- Unchanged
- Merge ready
- No
- Baseline priority
- Not loaded
- Revised priority
- Not loaded
- Relationship
- Not loaded
- Resolved findings
- 0
- Added findings
- 0
- Changed files delta
- 0
- Diff lines delta
- 0
- Project tests
- Not run
Review priority is deterministic triage, not a code-quality score. Human approval and repository project tests remain required.
- Provider / policy
- Runrail policy
- Basis / confidence
- Deterministic
- Authority
- None
- Browser network
- Disabled
The receipt is matched to the loaded audit checksum. It cannot execute a command, run a test, change a patch, approve a merge, or apply code.
The advisory receipt is invalid or does not belong to a loaded audit.
No audit loaded. Nothing has been uploaded or automatically stored. Saving a handoff is an explicit local download.
What the workspace does not do
It reads evidence. It does not run code.
The workspace cannot authenticate audit origin, open a repository, call any model provider, execute project tests, determine semantic correctness, perform a security review, apply a patch, or approve a merge. Generate audits and candidate-advisory receipts with Runrail Local, inspect them here, and run your repository's approved checks before deciding. Choosing Save review handoff explicitly writes a reduced JSON file to your download location.