Changelog

A new Under the hood page explains what happens when you press Scan and how SlopZero compares to other tools.

Look under the hood

• How SlopZero maps your app, passes precise code trees to decision models for judgment, and delivers verified remediation to your agent or editor.

On Starter, Pro, and Max, a GitHub repo can scan itself on every push to its default branch, daily, or weekly.

• Turn it on from the project page after your first scan. Each run uses one scan from your plan.
• Autoscan skips a commit that was already scanned, so it never spends a scan on the same code twice.

Tell Claude Code, Codex, Cursor, Grok Bot or Muse “Scan with slopzero.dev”.

Try it

• Sign up with Stripe Link: no form, no password. Have an account? Pick GitHub when your agent asks.
• It scans the folder you’re in, unpushed changes included.
• Out of scans? Approve a pack or a monthly plan in Link.
• It asks before fixing anything, then rescans.
• Link sign-up is US and Canada only for now; elsewhere, use GitHub.
• Something not right? Tell your agent: it sends your words to the SlopZero team and shows you our reply.

The top of the scan page now shows when we start checking your findings, without a reload.

• Before, it could keep saying your scan was waiting in line after the check had already begun.

Add SlopZero as a remote MCP server in Claude, ChatGPT, Cursor, and other clients. Your agent can start a scan of a pushed commit and bring the findings back to fix.

How to connect

• This release is MCP only — there is no CLI to install. Paste the server URL in your client; it opens a browser so you can approve. No API key.
• Works with Claude (claude.ai, Desktop, and Claude Code), ChatGPT and Codex, Cursor (IDE, web, and Cloud Agents), Replit, Lovable, Devin Desktop (formerly Windsurf), and VS Code / GitHub Copilot.
• Hosted clients need this public HTTPS site. localhost will not work from Claude, ChatGPT, Replit, Lovable, or Cursor Cloud Agents.
• Scans need the SlopZero GitHub App on the repo. The agent sends an install link if it is missing. Review or revoke the connection later under Account → Connected agents. Connections and revocations also show up in Activity.

A new review engine we're testing, built to flag fewer false alarms and finish review much faster.

Turn it on

• Choose Experimental under Account → Scan settings. It applies to scans you start after saving.
• Experimental scans are labeled on the scan page and while they run.
• Switch back to Standard any time.

While a scan is reviewing findings, the progress card can show a remaining-time note such as “about a minute left” instead of leaving you to guess.

• The estimate updates as review continues, not as a ticking countdown.
• Waiting in the review queue still says waiting, without a clock.

The dashboard loads faster when you come back, and project, scan, and triage pages respond quicker as you click through — with repositories ordered by recent activity and easier filtering on long lists.

Open dashboard

• Return visits and repeat project or scan views spend less time waiting when your repository list was loaded recently, including organization repositories.
• Repositories are ordered by recent activity, with last-push dates on each row.
• Filter by name when you have many repos — archived projects stay in the list too.
• Adding or removing a repository on the GitHub App updates those pages on your next visit.

When review finishes, the scan page, downloads, project counts, and Activity show the issues that were kept.

• Dismissed and not-sure rows stay off the scan page and downloads.
• Counts and severity on the project list match what stayed after review.
• Downloads use a shorter summary; if nothing was kept, the scan and export say so clearly.

Waiting review no longer looks active, and long live previews are easier to move through.

• Review waiting for its turn is labeled waiting, not running.
• Large live previews add previous, next, and a position slider.

Claim one bonus scan credit each UTC week when you share a finished scan on X. The same personal ?ref= link on Account works for new signups too — those are separate bonuses.

Open Account settings

• Share on X from a finished scan opens a pre-filled post with your personal ?ref= link — generic copy only, not your repo or findings.
• After you post, claim one free scan credit each UTC week from the scan page. You do not need a new signup for this bonus.
• Copy your ?ref= link anytime from Account settings — it uses your GitHub username.
• When someone new signs up with GitHub through your link, you can each earn one bonus scan credit.*

* Your GitHub account must be at least 7 days old when you sign up through a referral link.

See scans that are still queued or running on your Activity page without opening each project.

• Activity shows an Active scans section when you have work in progress.
• Each row links to the scan progress page for that repo.

The home page now explains what SlopZero does, walks through a scan, and shows pricing before you sign in.

View the home page

• Features, a four-step walkthrough, plans with one-time credits, and a plain-English FAQ — all public.
• Straight about the limits: which stacks work, what the double-check sends where, and that it isn’t a security audit.
• The FAQ now answers how long a scan takes and what happens to your code afterwards.
• Legal pages, the waitlist, and email sign-in share the same look.
• Switch between light and dark mode from the header; your choice carries into the app.
• The logo, browser-tab icon, and link previews now match the brand kit, on the home page and in the app.

The home page highlights our Product Hunt launch with bonus scans and a paid-plan discount.

• Sign up from Product Hunt for 4 bonus scans on top of the free monthly scan.
• Use promo code PH50 at checkout for 50% off paid plans for three months.

New users see a short dashboard checklist after signing in with GitHub.

• The dashboard walks you through installing the GitHub App, granting repos, running your first scan, and opening the results.
• Hide the checklist any time — your progress is tracked automatically.

Choose when SlopZero emails you about finished scans — per scan on the progress card, or as account-wide defaults.

• While a scan runs, toggle Email me when this scan finishes on the progress card.
• Account settings control the default for new scans and whether unopened-results reminders are sent.

Finished scans you have not opened yet show up on your dashboard so you can jump straight to the results.

• The dashboard lists successful scans you have not viewed since they finished.
• Each row links to the scan page; opening it clears the item from the list.

Get an email when your scan report is ready, plus gentle reminders if you have not opened the results yet.

• Report-ready email links straight to your scan page when a customer scan finishes.
• If you have not opened the results after 3 days, SlopZero sends one reminder; a final-day reminder goes out on day 6 (before history expires on Free).
• Opening the scan page stops further reminders for that scan.

On your first scan of a repository you can choose whether to include it in training. Turn training off for your whole account from Account settings.

• First scan of each repo asks whether to include it in training (default on). Later scans on that repo use your choice.
• Account → Scan training turns training off for every repo and hides the first-scan prompt.
• Change a decided repo from its project page.

While review is still running, the scanning card stays up and rotates through open findings plus quick context inside the card — not a full report.

• The progress card stays visible until review finishes — no full report shell while the scan is still running.
• Inside the card, a live preview rotates through open findings and a compact concern peek (stack, fit, watch surfaces, notes).
• When review completes, the judged report uses the same concern peek instead of the four-column map.

Signed-in Help can use PostHog Support. You can still open Help if you Decline optional analytics cookies.

• Accept still controls analytics cookies, replay, autocapture, surveys, and pageviews.
• When Help is on, a signed-in browser may keep a Help chat session in local storage. Decline does not turn Help off.
• A new Help ticket includes the page you started the conversation from. Follow-up messages can include a session id and a session-replay link.
• Help can use PostHog's widget or SlopZero's chrome on the same tickets.

When review is part of the scan, the progress card stays up until that review finishes. The report you see is the final list — not first-pass findings with kept / not-an-issue labels.

• The scanning card stays on screen until analysis and review are both done.
• The finished report does not show kept labels, a Not an issue section, or Previously dismissed items.
• Headline counts follow the open findings, not every first-pass row.
• Copy for LLM asks the model to work from Findings and treat other sections as context.
• Click the scanning card on a project to open that scan.
• If review is not part of the scan, first-pass results stay on screen with a compact review banner.
• If review is still pending after a timeout, the first-pass list says Review still pending.
• Copy for LLM stays hidden while that review is still pending, and the downloaded report says the list is first-pass.

When analysis has already finished, the scan page shows those results instead of staying on the first spinner. A later review still overlays the results until it finishes or times out.

• A finished scan leaves the first progress spinner even if the review feed is briefly unavailable.
• Review still shows on top of first-pass results while it is confirmed to be running.
• A brief review-feed outage keeps the review spinner when the last progress snapshot is still available.
• If the review feed cannot be reached, the page keeps checking for a finished review instead of waiting forever.
• After review finishes, checking for a stuck result uses the 10-minute live-updates limit, not the 30-minute review cap.
• A report that is still running a review job is left running so a later harness retry is not overwritten.

A scan is not finished when the first analysis pass says done. The progress spinner stays up through the second review pass. If that review cannot finish, the scan still keeps the first-pass findings.

• The spinner stays on the scan page until review finishes, including on top of first-pass results.
• Short review updates from the analysis service can appear under the spinner.
• A review outage no longer fails the whole scan or removes the workspace.
• You cannot start a new scan on the same project until the current review finishes.

Signup can stay invite-only, open a waitlist, or let anyone in. Invite links still grant that code's credits.

• Sign in with GitHub to join the waitlist if you do not have an invite yet.
• If you already have a code, enter it on the home or waitlist page — invite links do not redeem themselves while you are signed in.
• When signup is open, you can sign in with GitHub without a code and still redeem an invite for that code's credits.
• If you are already signed in, a missing invite says to enter a code — not that no account was found.
• Waitlisted accounts, or signed-in users still entering a code on home, that already have a billing customer can open the Stripe portal.

The home page and browser tab now say Ship vibe-coded apps with confidence.

• The line above the home headline is the new tagline.
• The default browser tab title is SlopZero – Ship vibe-coded apps with confidence.

The cookie prompt is one short line. Accept, Decline, and Privacy are unchanged.

• The banner now says we use optional cookies for product analytics.
• What Accept records, and that server events can still fire after Decline, stay on the Privacy page.

The home page now says what SlopZero ranks first, and the sample scan lists findings worst-first.

• The guest hero and signed-in invite page share one sentence about what could break and what to fix first.
• Benefit bullets cover missed problems, ranked worst first, and looping in an expert.
• The sample scan lists four findings by severity instead of a false-alarm tally.

When a scan or judgment cannot finish, you see a short retry message instead of internal analysis text.

• Failed and incomplete scans no longer show harness or worker error strings on the scan page, project list, or Activity.
• A failed scan with nothing to show uses one Scan failed banner, not a second empty-state card.
• Retry loading results and live failed progress stay on catalog copy if the analysis service reports an internal error.
• Downloaded reports use the same generic judgment-failed line.

The cookie banner and Privacy page now say exactly what optional analytics records after you Accept, including replay guards and digest-only errors.

• Accept still loads page views, allowlisted clicks, session replay, digest-only errors (PostHog may attach a capture-site stack), and surveys.
• Magic-link, share, and transfer pages stay out of replay, and finding titles are not autocaptured.
• Privacy names GitHub login and invite labels on server events, which can still fire after Decline.

Reload and What’s new sit with the app header while you scroll. When a deploy generation is set, reload waits for a newer generation during a rollout.

• The banners and header stick together, so scrolling no longer covers Home, breadcrumbs, or sign-out.
• Reload treats an older replica as old when we have a deploy generation — not just a different build id.
• Reload confirms a new version on a later poll, so one leftover replica cannot show the bar by itself.
• Activity, dashboard, and loading pages use the same wide column as Home, so the feed is not pinched on a large screen.

Reload and What’s new stay out of the way, and the public Terms and Privacy pages cover how the product actually works.

• The reload bar snoozes instead of disappearing, and both banners sit in the page instead of covering the header.
• Public labels use the same muted token as the footer so small type meets contrast requirements.
• Privacy and Terms describe redeemed invites, Calendar attendees, and that judgment can use a language model even when synthesis is off.

Reload when we ship, and catch up on changes when you come back.

• A tiny banner asks you to reload after a new version is deployed.
• Coming back after a public update shows what’s new — Changelog, Terms, and Privacy live in the footer.