Proof, not a status light
Every run keeps the screenshot and the exact lines Smith read on the page. You see what broke, not just that something did.
Smith opens your product like a customer would — in a real browser, on a schedule — and tells you the exact page that stopped working, with the screenshot to prove it. Before someone emails you about it. And when you want a second pair of eyes, ask it to look at any page and tell you how it reads.
What the monitor actually does
Checkout is failing
yourproduct.com/checkout
Smith saw: the page loaded, but the pay button was missing and the cart was empty.
Screenshot: attached to the run.
Next run: in 15 minutes.
Not a dashboard you have to read, and not a bot that guesses. Smith does the boring thing a customer would do — opens the page, looks at it — and tells you the moment it stops being right.
Paste a URL and say what working looks like, in one sentence.
A real browser opens it, screenshots it and judges it against your sentence.
An incident opens with the page, what Smith saw and the proof.
Every run keeps the screenshot and the exact lines Smith read on the page. You see what broke, not just that something did.
Uptime per check, every run in order, and incidents that close themselves the moment the page is healthy again.
No false alarms by design. Smith reports only what it actually saw on the page, and every incident carries the screenshot from the run that failed.
Uptime is the floor, not the whole job. Ask Smith to open a page and judge it — the layout, the copy, the flow, the thing that looks unfinished. It opens the real page in a real browser and answers from what it actually saw.
you · Open my pricing page and tell me if the plans make sense.
you · Read my signup copy out loud — where does it lose the reader?
you · How does the checkout look on a phone? Show me what you saw.
you · Did today's deploy change the hero? Compare it to yesterday.
Every answer comes with the screenshot Smith took, so you can see exactly what it judged. No "looks good to me" without evidence.
Opens the page itself
Quotes what is actually on screen
Connect accounts with OAuth in a couple of clicks, then point Smith at any MCP server you host yourself. Permissions are per tool, and you can revoke them at any time.
Model Context Protocol
Any MCP endpoint — HTTP or SSE — with exactly the tools you enable. Smith lists what it can reach before it uses it.
Every plan runs real browsers on a schedule and keeps the evidence. Pick how often Smith should look, and how far back you need to see.
$0forever
1 check · every 6 hours
About 120 runs a month
Enough to watch the one page that matters most.
Best for · Trying Smith on a single page
$19per month
10 checks · hourly
≈ 7,200 runs a month
For one person watching their whole product.
Best for · A solo builder shipping daily
$79per month
50 checks · every 5 min
Unattended, all day
The full watch, on every flow that earns money.
Best for · Monitoring that never sleeps
$29per user / month
Shared check pool
Pooled across the team
One workspace for everyone who gets paged.
Best for · Small teams on call
1 page, hourly
720 runs / month
1 flow, every 15 min
2,880 runs / month
Screenshot storage
90 days on Pro
Every run draws from your daily credits — 1 for a page, 2 for a journey — and chat uses the same pool. When a day's credits are gone Smith waits for midnight rather than billing you more.