For design studios & agencies — a system you own, not a subscription
The machine checks every file — rules, and an AI reviewer.
Your client approves from one link.
The record can’t be rewritten.
We build it into your studio’s workflow — tuned to your brand rules, wired into your tools, owned by you.
built for your business · $10 test budget on our infrastructure
Deployed & real — the checking machine is resting; recorded real runs play below
An invite lets you run it on your own work, on a test budget we fund.
1 · the machine catches it — before the client
Spelling on the artwork — possible spelling issues: SPECAIL
Recorded real run — Mimosa special — Instagram story (EN)
In short
Already running — deployed and checking real files since July 2026
Built for you — tuned to your brand, wired into your workflow, owned by you
Try it for real — invites carry a funded test budget; you never pay in the app
Watch it run — recorded from the live system, nothing staged
48 in · 3 caught before the client
Recorded real run · nothing staged
A whole campaign — 48 files in one drop, checked in minutes. The flagged tiles are the catches.
Step 1 of 8 · Batch · 0:00 / 0:38
Watch a real check — recorded from the live system, nothing staged
file read: PNG · 3.6 MB — within the limits this client confirmed
These findings are not an illustration — they are the machine's logged verdict on this file, drawn where it found them. Switch its eyes on and off:
This recorded check is the exact-rules set — size, file type, spelling on the artwork, brand colors. The AI reviewer is the other half, and it runs on drops that ask for it.
The numbers — sourced and measured
The email way — independent surveys
This machine — measured on the live system, on our own runs so far
The blocks — what a build for your studio includes
A build starts from the working core and adds the pieces your studio needs — each tuned to your rules during the engagement.
Working today — in the deployed system
Drop a whole campaign of files and say the goal in your own words. What to check and who approves comes back proposed — by the AI when it can be reached, from the files themselves when it can’t — and nothing exists until you confirm it, card by card.
Before your client sees the work, the actual files are read and checked two ways: exact rules — right size, right file type, spelling on the artwork, your brand colors — and an AI reviewer that looks at the work against your brief. Honest results per check, each saying which found it — never a guess.
Your client opens one link — no account, no password — points at the exact spot on the work, comments, and approves or asks for changes.
A misspelling was flagged on the light table, fixed, and the client only ever saw the corrected draft. Every new draft goes to the same link, and an onion-skin overlay blends the previous one under it, so your client sees exactly what moved.
A point, a box, a highlighter, an arrow, or a line drawn freehand — right on the work, wherever the problem is. Your client can attach a picture of what they mean, and all of it leaves as a checklist you paste into whatever you already run.
Every draft, mark, and decision lands in a record the database itself refuses to change. Who approved what, on which date, with the time it took — settled.
Drop your client’s logo once and every piece is checked for it — found at half-size to triple-size, anywhere on the artwork, with the spot marked. Say “they need this by Friday noon” in your own words and the date is read out of the sentence and counted down beside the work.
A campaign is dozens of files, not one. Drop them together, confirm once, and read one plain-words rollup: files checked, problems caught before your client, time to approval.
Two drafts next to each other with the differences highlighted and a slider between them — beyond the onion-skin.
Art director first, account lead second, client last — staged sign-offs per deliverable, each on its own link and each on the record. The imprimatur is granted only when the last one approves.
Brochures and decks reviewed page by page — the pre-check reads every page, a mark lands on page 4, and the same record holds it all.
Built onto it for you — on request
Marks pinned to the exact second of the cut. “At 0:14, hold the shot” beats a paragraph in an email.
Approvals posted to your team chat, finals pushed to your shared drive, every decision flowing into the systems you already run.
Your clients review under your studio’s name and look — the machine is yours, invisibly.
Fonts, logo clear-space and proportion, the rules your brand book spells out — tuned to your own guidelines, not a generic checklist.
Against the rented tools — from a first-hand teardown
The rented tools — Filestage · ReviewStudio · Frame.io · Ziflow
The build you own
The record
The audit trail is a paid add-on: a generated PDF at €329/month, live logs quote-only — or a feed that can be edited after the fact.
The record is append-only at every tier, enforced by the database itself. Corrections are new entries; history never rewrites.
The AI check
AI checking is “coming soon,” a chat box that never looks at the file, or ~$1 per check on top of the subscription.
The pre-check runs today, on the real bytes, against your own brand rules — at a per-file cost near zero.
Your numbers
Your approval speed is their sales pitch (“30% faster!”) while your own numbers sit behind “Get in touch.”
Your cycle times are computed from your own record and shown to you by default. Your numbers are yours.
Ownership
A per-seat tenancy you rent forever — roughly $4,000–6,000 a year for a 20-person studio, and it is never yours.
A build tuned to your rules and wired into your workflow, owned by you. One engagement, near-zero running cost.
July 2026. Filestage, ReviewStudio and Frame.io were operated from the inside — real accounts, real clicks, screenshots. Ziflow’s signup is walled behind a bot check we could not pass, so what it contributes here comes from its own published pages. The receipts live in the project record.
Yours — including the AI
The bespoke build gets better the more you use it: your corrections teach it your rules, your record grows into your proof, and the whole machine is yours, not rented.
Cheap, capable models — checking a file costs next to nothing. The deployed system you’re looking at runs on this tier today.
The receipts
The raw readout, for the curious
Size — passed · 14844 ms · api/claude-opus
File type — passed · 14844 ms · api/claude-opus
Spelling on the artwork — passed · 14844 ms · api/claude-opus
AI review — did not pass · 14844 ms · api/claude-opus
Every number above was computed from logged events on the deployed system — never typed in by hand. When you run your own project, your numbers come from your own record the same way.
A whole campaign at once — files in, plus one sentence in your own words about what matters and who approves.
The path — ask, try, commission
You never pay in the app; a commissioned build is a separate written agreement.
