On-brand product UI.
Even when AI
writes it.

One system of record for your product UI, built into your product in 90 days, with three consumers reading from it. A Figma library with 30-40 core components. The same components as tested code in your repo (React or Angular). And the Constraint Framework: machine-readable tokens, rules files, a design-system MCP server and CI gates, so Cursor, Claude Code and Codex build from your system instead of guessing.

Plus the proof layer: the Adherence Score, a dashboard that shows, PR by PR, how much of your UI ships from the system. Run by the two founders. No juniors, no long discovery, no agency markup.

90 days, fixed price, from €125,000. 95% of projects land between €125k and €185k. Merging many brands or carving out a system after an acquisition costs more.

50% on signing, 30% at week six, 20% at week twelve. No hourly billing, no day rates, no change orders under €5,000 of new work.

VPs of Design, CPOs and the CTOs they build with, at European scale-ups making €30M to €250M a year. Series B to PE-owned, 50 to 500 on the product team. Design usually feels the pain first; engineering owns the budget and the metric we guarantee — we're built to be bought by both.

Teams past the size where consistency holds by luck, shipping AI-made UI faster than they can review it, who want it built, not workshopped.

Not for: startups under fifty (use shadcn/ui, come back at Series B), or anyone shopping for hourly billing and weekly status decks.

Two firm guarantees, written into the contract.

Early exit at week two. If the direction isn't right, walk away with a full refund minus a flat €12,500 fee. You keep the week-two work.

30% faster or we work free. At week eleven we re-measure PR cycle time on your own GitHub data against the four-week baseline signed at kickoff — a developer survey runs alongside as context, never as the contract. If speed hasn't risen 30%, we work free until it does, up to 60 more days.

Miss again at day 150 and we refund the final 20%. You keep every Figma file, every line of code, every governance doc.

Two weeks before you commit to the Install. We prototype the token setup on your brand, scope exactly what we'd build, and hand you a one-page summary your CFO can sign off — plus the vendor pack procurement will ask for, so supplier onboarding runs while you decide instead of after.

The work is yours either way. Hand it to your own hire or another vendor, and we'll do a free one-hour handover call.

The full €9,500 comes off the Install if you book within 30 days. Free if you go ahead, yours to keep if you don't.

Yes — and not by prompting alone. We make your system machine-readable and then enforce it: design tokens as one source, per-component metadata, rules files for Cursor, Claude Code and Codex, a design-system MCP server your tools query, and CI gates that catch drift before it merges.

Here's the uncomfortable truth: your AI already has a design system — it's shadcn, and it isn't yours. Generic, React-shaped, off-brand. We replace that default with your system, and the Adherence Score shows it holding, PR by PR.

In-house means a real team once you count design, engineering and the pipeline work — around €500k a year loaded for four seniors. Six months to hire, nine to ramp, and people stay 24 months on average. Most internal systems teams also report being understaffed within a year.

The cheap routes stay cheap by leaving the job unfinished: a €30-60k build hands you Figma files your AI can't read, and shadcn is free because it's everyone's design system. Like-for-like — design plus tested production code plus the AI layer — agencies run €300-400k over a year, a Big 4 firm around €600,000.

Ours is 90 days, from €125,000, with two firm guarantees. Senior founders on every piece of work.

Up to you. The Monthly is €9,800/month for past Install clients only, capped at six clients at a time. Both founders on every piece of work.

It covers what actually rots: new components, token updates, code-side fixes, and quarterly re-runs of the agent eval — AI models and tools change fast, and the rules that keep them on-brand need re-testing when they do.

Mostly async, one job at a time. The 4-hour rule: we reply on Slack within four working hours, or the month is free. That's in the contract too.