Secrets in code
API keys, service keys and passwords committed to the repo, including old commits.
Clones the repo read-only and scans files and the last 300 commits. Custom.
Seven security checks for apps built with Lovable, Replit or Bolt. Plain-English findings with a fix prompt for your builder.
See a sample reportBuilt for apps made with
Non-engineers now ship apps that hold real customer data. Scanners find the first holes; nobody owns the rest.
of 62 Lovable app scans by an independent researcher showed critical or high-severity findings. 77% of the apps ran on Supabase.
I scanned over 50 Lovable apps for security vulnerabilities, Mediumvulnerabilities and 400 exposed secrets found across 5,600 AI-built apps, per a study cited in the same write-up.
Cited in the same Medium write-upof 5,000+ apps on Lovable, Replit, Base44 and Netlify that lacked basic security leaked private data, including medical and financial records.
Freevacy, summarizing WIRED's report on the RedAccess studyThis is the report the Full Scan delivers, playing its sample. Seven checks, a health score, every finding with its fix prompt.
A Full Scan runs all seven in parallel and merges the results, so one issue found by two tools appears once.
API keys, service keys and passwords committed to the repo, including old commits.
Clones the repo read-only and scans files and the last 300 commits. Custom.
Keys shipped to the browser, such as a Supabase service key.
Downloads the live JavaScript bundles and matches known key patterns. Custom.
Tables without row-level security, loose policies, exposed sensitive columns.
Reads the database catalog with the read-only role, never your rows. Custom.
Tables anyone can read with the app's public key.
Uses the public key from the bundle to request one row from each exposed table. Never writes or deletes. Custom.
Missing security headers, insecure cookies, information leaks.
Reads the live pages without attacking them: cookies, mixed content, public source maps and exposed files. Custom.
Missing basic headers, a certificate close to expiry.
One request to the home page. Custom.
Packages with known vulnerabilities.
Reads the lockfile and asks the OSV database about every version. OSV.dev.
The report is the product, so most of the design went into it. Every page answers one question an owner has.
LeakScan does not touch your app. It writes the instruction, and your builder makes the change while you watch.
customers, bookings and invoices answer to the public key with no row-level security.
LeakScan never applies a fix itself. Every scan is read-only, so there is no need for write access, pull requests or SQL on your database.
Pay at checkout, connect the repository with a read-only token, run one SQL snippet that creates a read-only database role, and paste the connection string.
Prove you control the domain with a DNS record or a file on the site. No check runs before this.
A worker runs all seven checks in parallel and merges the results, so one issue found by two tools appears once.
LeakScan ranks every finding by severity, explains it in plain English, and writes a fix prompt you can paste into your builder.
You get an email with a link to the report page and a PDF: a health score, the findings in order of severity, and the fix prompts.
Within 14 days, one click reruns the checks and marks each finding fixed or still open.
After the rescan or at 14 days, LeakScan deletes the stored credentials. The report gives you the one line of SQL that drops the database role and a reminder to revoke the GitHub token.
If your app has real users and holds their data, this is for you. You do not need to know what row-level security is.
A leak or a lost database hits the business directly, and there is no developer on staff to catch it.
Revenue depends on the app staying up, and one breach is enough to lose the trust you earned.
Clients ask what happens after launch. A scan report under your brand is the answer.
The quick test: the app has real users, it stores personal or payment data, and you cannot say whether row-level security is on.
Lovable runs a security check on every publish and offers a one-click AI review. Good for catching issues while you build, in the builder's own terms.
Priced at $9 to $49 a month, they hand you a list of findings from the outside. The fixing, and the database, are still on you.
One price per app. It reads your database with a role you control, probes what the public key exposes, explains every finding in plain English and writes the fix prompt you paste into your builder.
Every check reads. Nothing writes. Here is exactly what LeakScan can and cannot touch.
A GitHub token that can read your repo's contents and nothing else.
One SQL snippet makes a read-only role. The report includes the one line that drops it.
Your live app is scanned the way a visitor sees it, and only after you verify the domain.
Encrypted at rest, fetched only by the job that needs them, deleted after the rescan or at 14 days. You can delete them earlier with one click.
Each scan runs in a fresh container, and the report lists every action LeakScan took on your app.
Every check, the full report with fix prompts, one free rescan.
No plans, no seats, no monthly fee. Pay once per app, get the report, fix what it found, and rescan for free within 14 days.
Launch pricing, shown in USD. The price may change before the public launch on November 22, 2026.
These dates come from the build plan and move only if the gates before them slip.
The seven checks run on the first apps while every report is read by hand. Every gap becomes a new check.
All seven checks with the report and one free rescan.
Terms, privacy policy and an outside security review of LeakScan itself, done before the doors open.
Firebase, Neon, Replit-hosted apps and Bitbucket, in the order customers ask for them.
Paste your app's address to start a Full Scan. Read-only, one price, report with fix prompts.