What FlowPatrol does and what it costs

FlowPatrol checks the signup, login, and checkout paths on web apps people ship with Lovable, Bolt, v0, Replit, Cursor, Claude Code, and similar editors. When a path breaks, it emails the failing step, a screenshot, and a prompt you can paste back into the editor.

Anyone can run a free scan of one public URL from the homepage. That scan loads the page. It reports HTTP status, HTTPS, load time, console errors, broken images, a sample of up to 12 links, failed network responses, and a screenshot. It does not click through a form, and it does not keep watching the app.

A paid plan stores flows you describe in plain English and reruns them on a schedule, from Run now, or from a deploy webhook. A step can open a same-host URL, click visible text, fill a field by label, placeholder, or textbox name, and assert that the page shows text or that the URL contains a path. Test emails end in @example.com.

It is for a founder or builder who changes a web app by prompting an AI editor and needs signup, login, and checkout to keep working afterward. The usual way those breaks show up is a user who cannot create an account, or a payment that never starts.

It is a poor fit if you need full browser coverage: many roles, an email inbox, captchas, passkeys, native mobile apps, or journeys that leave your hostname to finish on a card form. You may only point it at sites you are allowed to test. The terms say that, and the privacy policy says what a scan stores: the email you submit, the URLs, flow text, results, and screenshots.

The live pricing section is the source of truth. Free is $0. See what a scan finds. No ongoing watch. Starter is $19/mo. One app, watched every day. Pro is $49/mo. A few apps, checked as often as the platform allows.

Paid plans are PayPal subscriptions billed monthly in US dollars. Cancel in PayPal. The free scan does not start a subscription and does not watch the app after the report. Refund terms are on the refunds page: cancellation stops the next renewal, and the current month is not refunded except where the law requires it or a charge was in error.

You paste the public URL of the app. You describe each flow in a sentence, such as “Sign up with a new email, then open the dashboard.” FlowPatrol turns that into a step plan when an AI key is configured. With no AI key, the plan falls back to a page-health check: load the URL, record status, console errors, and a screenshot. Add a key later and regenerate the plan.

Starter includes 1 app and 3 flows, checked on a 24-hour interval. Pro includes 3 apps and 10 flows, with an interval of 6 hours where the host allows that cron. On Vercel Hobby, cron runs at most once a day, so Pro checks run daily until the project is on a plan that allows a more frequent cron. Run now and the deploy webhook are on both paid plans. Pro also sends a weekly health summary.

A failure email names the app, the flow, the step that stopped, and the error, and it includes a screenshot link plus a fix prompt. The prompt tells the editor which flow broke and asks it not to change unrelated parts of the app. Console errors are stored as a note. They do not, by themselves, fail a watched flow.

Manual testing is the right way to launch. The checklist on how to test a Lovable app before launch is that pass: a private window, a new email, login twice, a test-mode payment if you charge, and notes about button text and the success sentence. You can see confirmation emails, captchas, and hosted card pages. Manual testing has no stored run and no alert. The break after a later AI edit waits until you remember to click through again, or until a user tells you.

A generic uptime monitor requests a URL and alerts when the status is unhealthy or the request times out. That catches a site that is down. It does not catch a signup button that was renamed, a form whose label disappeared, or a checkout page that returns HTTP 200 with the pay button gone. FlowPatrol’s free scan is a single page-health look (status, HTTPS, console, links, images, screenshot), which is more than a ping and still does not complete a flow. Watched flows click and fill. An uptime check will not.

A Playwright suite you write yourself is the more precise tool. You choose locators, fixtures, and assertions, and you can cover an inbox, role combinations, and a hosted payment page if you wire that up. The maintenance cost is the same AI edit that renames a button or a label: the locator fails until you update the test. FlowPatrol asks for the flow in English and builds a short plan from the live page. It is narrower on purpose. It stays on your hostname, caps the plan at 12 steps, uses @example.com addresses, and skips captchas, passkeys, and email codes. If you already maintain Playwright, keep it for the cases it covers. Use a watched flow for the few customer paths you would otherwise retest by hand.