Ojo app by Buena, the creative studio for good causes

QA and monitoring for every site we ship: links, layout, SEO and a design score.

Client: Buena apps
Year: 2026
What we did: Website, App
Result: Running on 2 live client sites

By the numbers

17automated checks
2×daily crawls
2client sites monitored

Ojo watches a published website and reports what changed since the last check: drafts published by mistake, broken links, edited copy, defects that only show up on mobile. We built it so nobody has to check a client's site by hand.

Client sites break quietly, while they're still up

Framer and Webflow sites are edited by people without a technical background, and publishing is one button. So the expensive mistakes are the quiet ones: a draft goes live, a link dies, copy is fixed on desktop and left broken on mobile.

Links rot long after launch. On one client site, none of the broken links we found had broken on the day they were published. They broke later, while nobody was watching.

Uptime tools miss all of it, because the site never goes down. SEO crawlers run when someone remembers, and hand back a 400-row audit with no memory of the last one.

Not "is the site OK?" but "what changed since yesterday?"

Ask a different question: what changed since yesterday, and did somebody mean to change it? That is the version of the question a person can act on.

Twice a day Ojo crawls the live site and compares it with the previous run. Every finding is tagged with the section it lives in and the date it first appeared, so "17 broken links" becomes "this one broke on August 31".

Mark a finding as intentional and Ojo keeps checking it. If the content underneath changes, the mark expires, so an accepted finding never becomes a place for a new problem to hide.

17 checks with no AI guessing, so every alert can be trusted

Seventeen checks in six groups: publishing, availability, links, content, responsive layouts and staging versus live. All of them are deterministic, built on diffs, hashes and status codes, so every finding can be traced to two snapshots.

A monitoring tool that cries wolf once gets ignored forever, so Ojo is built to report less rather than report wrong. No link is called broken on the first try, empty runs are thrown away, and sites that simply block bots are not flagged.

On top sit a traffic-light dashboard, a design score and link repair. For a broken link, Ojo finds and verifies working candidates first; a model may only choose among them, never invent a URL.

Deterministic checks
Diffs, hashes and status codes. No model decides what is broken.
Since when
Every finding carries the date it first appeared.
Verified fixes
Replacement links are suggested only after they are checked to work.

Running twice a day on two live client sites

Ojo runs in production on two of our client sites, Sphere and CareSpace. It also became the single publish pipeline for oursphere.org.

Detection costs nothing in model spend, because no model is involved. Only link repair calls one, and our spec puts that at well under a dollar a month per site.

It means nobody on the team has to check a client's site by hand. When the tray icon turns red, there is something real to fix.