By Maciej Pyrc, who maintains Hype Stack.Published .
v0 is very good at one job: a screen from a prompt or a screenshot. People look for an alternative when the job turned out to be the app.
Vercel's v0 generates React with Tailwind and shadcn/ui from a description or a picture, and the output looks designed. It has grown toward whole Next.js projects, but the screen is still what it does best, and that is what most people typing v0 alternative have found out: they wanted an app with a database and a login, or they wanted the output in a codebase that is not Next.js on Vercel, or they wanted to stop paying per generation for something a coding agent in their own repo could do.
So the alternatives split by which of those three you meant. This page takes them in that order, and the last one is ours, said plainly.
Sort out which alternative you mean
What did v0 stop doing for you?
Lovable, Bolt or Replit build the application; v0 never claimed to.
- Lovable for a non-technical founder who wants a URL with a login tonight
- Bolt for a developer who wants the files and a terminal in the browser
- Replit for one tab that hosts, stores and deploys
A coding agent with shadcn/ui installed does v0's job inside your repository.
- Claude Code, Cursor or Codex generate the same components v0 does, against your real data
- Dyad runs the same loop locally and open source; Onlook edits React visually in your repo
Figma's generation features, Builder.io or a prototyping tool, when the deliverable is a design.
- These produce mockups and visual layouts rather than a component you ship
- Right for a designer handing off; wrong for a developer wiring data
The second branch is where most developers land, and it is the one the listicles under-explain.
What v0 gets right, so the alternative has to match it
Before leaving, name what you are giving up. v0's output is standard React, Tailwind and shadcn/ui, which is why it drops into so many codebases. Its design taste is the best in the category. Its design mode lets you nudge spacing and copy without spending a generation. And it deploys to Vercel in a click, which for a Next.js project is the whole deployment story.
An alternative that produces worse-looking components, or components in a vocabulary your app does not use, is a step down even when it fixes the thing you left over. The test for any option below is: same primitives, same quality, in the place you actually work.
The builders, when the screen was never the problem
All three answer "I need the app". None of them answers "I need v0's screens inside the app I already have", and that is the more common reason to be on this page. Bolt vs Lovable and v0 vs Lovable go deeper on the two most people shortlist.
The developer's alternative: shadcn/ui plus a coding agent, in your repo
v0's trick is not magic. It knows shadcn/ui's components and Tailwind's classes and has seen a lot of good interfaces. A coding agent in a repository that already has shadcn/ui installed knows the same things, with one advantage v0 cannot have: it can read your routes, your types and your data, so the screen it writes is wired on the first pass instead of filled with mock data.
The loop that replaces v0 for a developer
Start from a repository with the primitives installed
shadcn/ui, Tailwind, a design token file and a layout shell. An empty folder makes the agent invent all four.
Give the agent the picture
Claude Code and Cursor both take a screenshot or a sketch. Describe the screen, name the data it shows, and ask for a plan before any file changes.
Wire it to real data in the same prompt
Point the agent at the SDK call or the route. The component arrives using real types, which is the step v0 output always needs afterwards.
Iterate in the editor
Spacing, copy and states are small follow-up prompts. Nothing is metered per generation; the cost is your agent subscription you already pay.
Dyad runs this loop as a local open-source app if you prefer a v0-shaped interface; Onlook does it as a visual editor over your React code.
Where Hype Stack lands
We are the repository in step one. Hype Stack is an MIT template with a React and TanStack Router frontend on Tailwind and shadcn/ui, the same vocabulary v0 speaks, a Hono backend on Postgres behind it, and packs for auth with organizations, Stripe billing and an admin app as source files. It ships rules files and an MCP server so Claude Code, Cursor or Codex work inside the conventions, and the frontend SDK is typed from the backend routes, which is what makes "wire it to real data in the same prompt" true rather than hopeful.
A v0 component also pastes in with almost no translation, so the honest shape for many teams is both: v0 for the one marketing page that has to look striking, the agent in the repo for the forty product screens that have to work.
What you get, and what you do not
The home for the screens
Routing, layouts, auth guards, settings and billing screens exist. The agent fills in the screens that are yours.
No generation meter
Nothing in the repository is metered. The cost is the coding agent you already use and your own hosting.
Not a design tool
There is no prompt-to-screen interface here. The agent is yours to bring, and v0 stays the better tool for a one-off design exploration.
What you write yourself is the product: the screens your customers pay for, the data behind them, the rules that make your app yours. The template is where they go so they keep working after the first week.
For the head-to-heads, see v0 vs Lovable and Bolt vs Lovable, and for the wider field, Best AI App Builder. The design side of building the screens yourself is in How to Design a Web App.