Design cleanup for vibecoded apps
You vibecoded a working product with Lovable, Bolt, v0, Cursor or Claude Code. Now it needs to look like your product instead of the tool’s defaults. We clean up the design on the code you already have: flows, missing states, one UI system, copy and mobile, plus brand and marketing site where you need them. No rewrite. One senior designer leads the work end to end, as part of Partner.
Vibecoded with Lovable, Bolt, v0 or Cursor? This is for you
This page is for founders and small teams with a vibecoded app, built with Lovable, Bolt, v0, Replit, Base44, Cursor or Claude Code. The app works. You may already have first users or a pilot. But:
- It has the same slop look as every other app made with the same tool.
- Screens drift from each other.
- Empty, loading and error states are missing.
- Mobile is rough, and people drop off during onboarding.
- Every new prompt undoes an earlier design fix, and the credits keep going.
It usually becomes urgent right before a launch, an investor demo, a first paid customer or a first design hire.
The fixes go into the product you already have, or into specs your AI tools can follow, so nothing gets lost between a design file and the code.
Why vibecoded apps look like slop
For a boring reason: with no design system to follow, AI coding tools fall back on their defaults. Each prompt solves one screen with whatever the tool assumes, so the app looks like every other app made with it, and slightly different on every page. The tools did their job and got you to a working product. To stay consistent, they need a written system they can read before every change, and that’s the part we build.
What we fix
We start with what a new user sees first and work outward.
Flows and first session
Sign-up, onboarding and the path to the first useful result. If people leave before they get any value, this is where we look first.
The states demos skip
Empty, loading, error and permission states. They’re easy to miss when you build screen by screen, and they’re where users decide whether the product is reliable. We design each one so the app still makes sense when there’s no data yet or something fails.
One UI system
Type, spacing, color and components defined once as tokens and applied across every screen, so the app stops drifting and the default look goes away.
Copy and mobile
Labels, buttons, empty states and error messages rewritten in plain language. Layouts checked and fixed on real phone sizes.
Brand and marketing site, where you need them
If the identity or the landing page doesn’t match the product, we fix them as part of the same system: identity rules, type, color, voice and a marketing site that looks like the app it sells. If you only need the brand or the site, brand identity starts at US$6k and websites at US$10k. See services.
Rules your AI tools can follow
A cleanup only lasts if the next prompt respects it, otherwise the slop creeps back in. Along with the redesigned screens, you get:
- Design tokens for color, type, spacing and layout.
- Component specs, including every state.
- A
DESIGN.mdfile that describes the system in plain text your tools can read. For example, which button style is primary, how much space goes between sections and what an empty table shows. - A short note on how to prompt against it.
Point your CLAUDE.md, Cursor rules or Lovable project context at these files and the tool works inside the system instead of guessing. More on this in Where AI belongs in a design workflow.
How it works
- We use your app like a new user. We sign up and go through onboarding, on desktop and on a phone, noting where it breaks or gets confusing.
- We agree on the priority flow. One flow first, usually the path to the first useful result.
- We fix the system and the flow together. Tokens and components first, then the screens that use them, on your existing code or as specs your tools follow.
- We hand it back with rules. Your team and your AI tools keep building on the same system.
What it costs and how long it takes
The cleanup runs as one of our existing engagements:
- Launch, from US$12k. About 4 weeks, fixed scope. For when the product and brand need to be ready for launch.
- Partner, from US$6k/mo. One focus at a time, pause anytime. For when the app is live and needs ongoing polish.
Not sure which one fits? Send us the app and we’ll tell you.
Full details on Partner.
What we don’t do
Backend, databases and security reviews are out of scope. If we spot issues there, we flag them so a developer can take a look. Our work is design, UX, UI, brand and site.
Why FormForm
- Twenty years of creative direction and more than a decade running a studio.
- We build and ship our own SaaS products with AI coding tools, Claude included.
- One senior designer leads every project end to end.
- Based in Argentina, working with teams worldwide.
FAQ
No, in most cases. Cleaning up an AI-built app is a redesign without a rewrite: we work on the interface layer. Your logic and data stay where they are.
Yes. We work with Claude, mostly Claude Code, in the project you built the app with. You also get design tokens, markdown files (like a DESIGN.md) and a development environment, so your own tools can keep building on the same system.
Yes, where the product needs it. They’re part of the same system as the app.
Keep the DESIGN.md, tokens and component specs in the project and point your tool’s instruction file at them. We set this up as part of the handoff.
They’re out of scope. We flag anything we find for a developer.
Launch starts at US$12k and takes about 4 weeks. Partner starts at US$6k/mo, one focus at a time.
The short form on Partner: name, email, company or product, stage, what you’re after, budget, timeline and a message. A link to the app helps too.
Send your brief
Share a live link, your stage and your timeline. We reply within two working days, and work starts within one to two weeks.