When you ask v0 for a web app without naming a stack, what does it generate, and how reliable is it with another framework?
answer
- one framework, one component kit
- the host's own framework first
- typed React, utility CSS
- other frameworks: possible, less reliable
- nothing external until an integration
basics
~20 sv0 defaults to a Next.js app written in React and TypeScript, styled with Tailwind CSS and built from shadcn/ui components. It can use other frameworks, but its docs say Next.js gives the most reliable results.
solid answer
~40 sv0 generates a **Next.js** application by default: React with TypeScript, Tailwind CSS for styling and shadcn/ui for components, with backend endpoints written in App Router conventions. A screenshot upload is recreated with shadcn/ui components too, and AI features follow AI SDK 6 patterns. You can ask for another framework or component library and v0 will try, but its own pages say Next.js "provides the most reliable results". What the first generation does not include is anything outside the code: a database arrives only when you add an integration such as Neon, Supabase, Upstash or Vercel Blob, API keys belong in the project's environment variables, and a chat started without a project gets its Vercel project the first time you publish.
go deeper
Recall the default stack in one breath: Next.js, React, TypeScript, Tailwind CSS, shadcn/ui. Add that other frameworks work but less reliably.
Explain why the default matters: the sandbox dev server, Design mode's Tailwind awareness and Vercel deployment all assume it, so leaving it costs correction rounds.
Show you know what the generation leaves out — persistence, secrets, a repository, a deployment — and how each one is added before engineers inherit the app.
Frame the default as a lock-in trade: a team on another framework either teaches v0 its design system or accepts more rework on every generation.
## What v0 is and what it emits **v0** is Vercel's hosted prompt-to-app builder: you describe an app in a chat (in any language), attach a screenshot, paste a Figma link or import a repository, and v0 writes the code, runs it in a preview and deploys it to Vercel. When you do not name a stack, its output is predictable, and knowing the default tells you what you are about to own. | Layer | v0's default | Where the docs say so | |---|---|---| | Framework | **Next.js** (App Router conventions for backend endpoints) | full-stack apps page, FAQs | | Language | React with **TypeScript** | FAQs | | Styling | **Tailwind CSS** | FAQs, screenshots page | | Components | **shadcn/ui** | screenshots page: "v0 defaults to using shadcn/ui components" | | AI features | **AI SDK 6** patterns for chat, streaming and tool calling | FAQs | | Data access | Supabase, Neon or Upstash clients, **without ORMs by default** | full-stack apps page | ## Why the default is Next.js The full-stack apps page gives the reason in product terms: Next.js lets v0 **colocate frontend and backend code** through server actions and API routes, so one generation can hold a page and the endpoint it calls. The rest of v0 is built around that output: - the chat's preview runs in a **Vercel Sandbox** with a framework-aware dev server that detects Next.js, Vite and generic Node.js projects; - **Design mode** is Tailwind-aware and surfaces Tailwind-compatible values when it edits an element; - **Publish** deploys to Vercel, the platform the generated code is tuned for. How those frameworks work is their own subject; for v0 the point is that they are the path of least resistance. ## Asking for something else v0 does not refuse another stack. The docs say plainly that "while v0 can use other frameworks, Next.js provides the most reliable results", and the screenshots page adds that v0 "has knowledge of other component libraries and frameworks" but generates better results with Next.js, Tailwind CSS and shadcn/ui. Two practical consequences: 1. Expect more correction rounds when you push v0 off its default, because you are working against what it does best. 2. If your team already has a component library, the supported route is to teach v0 that design system (Design Systems 2.0 saves it as a skill in your workspace) rather than to restate it in every prompt. ## What the first generation still lacks A new v0 app is code plus a running preview. Several things only appear when you add them: - **Persistence**: no database exists until you connect one from **Project menu → Settings → Integrations** or ask v0 to add it; the integration provisions an account with the provider through the Vercel Marketplace and adds environment variables to the project. - **Secrets**: API keys for outside services go into the project's environment variables (the **Vars** panel), not into the prompt. Only variables prefixed `NEXT_PUBLIC_` reach browser code, and v0 warns about risky `NEXT_PUBLIC_` usage. - **A repository**: Git is optional; connecting GitHub creates a private repository that then becomes the source of truth for the code. - **A deployment**: a chat started without a project gets its Vercel project on the first **Publish**, and the app gets a stable production URL. ## Ways to start a generation - a text prompt (typed or dictated); - a screenshot or other attached file, which v0 analyses for layout, colours and components; - a Figma file or frame link (a paid-plan integration); - **Import from GitHub** or **Upload from computer** (a ZIP) from the **+** menu in the prompt bar. Whatever the input, the output lands in the same default stack unless you ask otherwise. ## Steering the output without leaving v0 Two v0 features change how a generation behaves without changing the stack: - **Instructions** are reusable prompts saved to your account and applied from the **+** button in the prompt bar. The built-in **Plan Mode** preset makes v0 write a plan of approach, steps and potential challenges and wait for approval before changing code; **Be Concise** shortens its replies. A custom instruction (a title plus a rule, such as "design for mobile screens first") applies across chats and projects while checked. - The **code viewer** (the **Code** tab) is read-only: you select lines, attach a comment and send it with your next prompt, so every change still goes through a generation. Both are useful when a team wants predictable output from a builder whose default it did not choose.
- Does a freshly generated v0 app come with a database?No. You add one from **Project menu → Settings → Integrations** or by asking v0, which opens the Vercel Marketplace with the provider's terms. Adding Neon, Supabase, Upstash or Vercel Blob provisions an account on that service and adds the environment variables the app needs. For SQL-based integrations v0 can then generate and execute SQL that creates, updates and drops tables.
- How do you make v0 propose a plan before it writes code for a large change?Apply the built-in **Plan Mode** instruction: open the **+** button in the prompt bar, check it under Instructions and send your prompt. v0 then writes a plan covering approach, steps and potential challenges and asks for approval before implementing. Checked instructions stay applied until you uncheck them; **Be Concise** is the other built-in preset.
- v0 only produces static mockups that cannot run server code
- v0 refuses any framework other than Next.js
- Every v0 app ships with a database already provisioned
- API keys should go into the prompt so v0 can wire them in
- v0 writes its own component library instead of shadcn/ui by default