v0 is an AI app builder and coding agent from Vercel that turns prompts, mockups and existing repositories into working web apps and deploys them on Vercel. It started as a prompt-to-interface tool and has since become a full agent with a sandbox, a Git workflow, MCP connections and a public API, so a lot of what is written about it describes an older product. This page sets out what v0 is today, what each plan costs, how its models and credits work, what the agent can do, which admin controls exist and how the product has changed. It is written for developers, designers, product managers and the people who approve tools for a team. Every fact comes from v0's own documentation, pricing page and Vercel's blog. Last checked: 2026-11-15.
This is an independent guide from NHI Mgmt Group, based on v0's public documentation as of November 2026. It is not sponsored or reviewed by Vercel. Product features change, so confirm settings against v0's current documentation.
Key takeaways
- v0 is now an agent, not just a UI generator. Vercel renamed v0.dev to v0.app on 11 August 2025 and rebuilt the product in February 2026 around a sandbox runtime that can import any GitHub repository.
- Four open plans. As of 15 November 2026 the pricing page lists Free, Plus ($30 per user per month), Business ($100 per user per month) and custom Enterprise pricing; the older Premium plan is being sunset and is closed to new users.
- Usage is credit based. Paid plans include $30 of monthly credits per user plus $2 of daily login credits, and four v0 models (Mini, Pro, Max, Max Fast) draw credits at published per-token rates.
- Git is built in. When a repository is connected, v0 works on a branch per chat, commits each change, opens pull requests and never pushes directly to main.
- Training depends on plan and settings. Enterprise content and content from teams that opt out are not used for model training; Business has the opt-out on by default.
What is v0?
v0's documentation describes it as an AI agent that helps people build real code, full-stack apps and agents from prompts. You describe an idea in a chat, or upload a wireframe or mockup, and v0 writes the code, runs it in a sandbox, shows a live preview, diagnoses its own errors and can deploy the result to Vercel's infrastructure. The FAQ calls it an AI-powered development platform that turns ideas into full-stack web apps through a chat interface, built by Vercel and designed to deploy on Vercel.
The audience is wider than developers. The docs list product, design, engineering, data, marketing, education and support teams as users, and Vercel's February 2026 announcement framed the rebuilt product as a way to move AI-generated apps "beyond prototypes" through normal Git review. The same post said more than 4 million people had used v0 since it became generally available in 2024.
v0 has four main surfaces: the web app at v0.app, an iOS app announced on 24 November 2025, a public API (the new version became generally available on 5 August 2026) and a v0 MCP server that lets other agents and IDEs create and manage v0 chats.
Key facts at a glance
As of 15 November 2026, from v0 and Vercel pages.
| Field | Value | Source page |
|---|---|---|
| Owner | Vercel | v0 FAQ |
| Product type | AI app builder and coding agent with sandbox, preview and deployment | v0 docs |
| Web address | v0.app (renamed from v0.dev on 11 August 2025) | Vercel blog |
| Surfaces | Web app, iOS app, v0 API (v2), v0 MCP server | Vercel blog, v0 docs |
| Default stack | Next.js, React, TypeScript, Tailwind CSS, shadcn/ui, AI SDK 6 | v0 FAQ |
| Free tier | Yes, 7 messages per day | Pricing page |
| Paid plans | Plus $30, Business $100 per user per month; Enterprise custom | Pricing page |
| Own models | v0 Mini, v0 Pro, v0 Max, v0 Max Fast | Pricing page |
| Code ownership | Vercel does not claim ownership of generated code | v0 FAQ |
| Training on customer content | Not for Enterprise or opted-out teams; opt-out on by default for Business | v0 FAQ, pricing page |
| Attestation | v0 is in scope of Vercel's SOC 2 Type 2 report (Security, Confidentiality, Availability) | v0 FAQ |
| Offline use | No, an internet connection is required | v0 FAQ |
Plans and pricing
The table copies the plan structure on v0.app/pricing as of 15 November 2026. Prices are in US dollars, per user per month, and exclude VAT, GST and other taxes.
| Plan | Price | Usage | Notable inclusions |
|---|---|---|---|
| Free | $0 | 7 messages per day | Deploy apps to Vercel, Design Mode visual editing, GitHub sync |
| Plus | $30 per user | $30 monthly credits per user, $2 daily login credits per user | Access to all models, extra credits shared across the team, centralised billing on vercel.com, shared chats and collaboration |
| Business | $100 per user | $30 monthly credits per user, $2 daily login credits per user | Everything in Plus, training opt-out on by default |
| Enterprise | Custom | Contract | Everything in Business, data never used for training, SAML SSO, role-based access control, priority access with no queues, guaranteed support SLAs |
Premium: the FAQ lists a fifth plan, Premium at $20 per month, and says it is being sunset and is not open to new users. The FAQ also states that the February 2026 product update did not change pricing.
Using a ChatGPT plan: v0's ChatGPT page states that ChatGPT Plus, Pro and Business users can build in v0 with their ChatGPT plan, so eligible responses count against that plan instead of v0 credits. A v0 credit balance is still needed, because each request carries a small platform fee and image generation, delegated subagents and supporting agent work always use credits. When the ChatGPT plan limit is reached, v0 pauses rather than switching to credits automatically.
Tracking spend: usage costs and token consumption are shown at v0.app/settings/usage. Additional credits can be bought outside the monthly allowance on paid plans.
Models available
v0 publishes per-token rates for its own models. Credits are drawn at these rates, per million tokens, as of 15 November 2026.
| Model | Input | Cache write | Cache read | Output |
|---|---|---|---|---|
| v0 Mini | $0.20 | $0.25 | $0.02 | $1.20 |
| v0 Pro | $2.00 | $2.50 | $0.20 | $10.00 |
| v0 Max | $5.00 | $6.25 | $0.50 | $25.00 |
| v0 Max Fast | $10.00 | $12.50 | $1.00 | $50.00 |
The pricing page does not state context windows, output limits or the base models behind each v0 model, so they are not given here.
Models through a ChatGPT plan: the ChatGPT page lists GPT-6 Astra, GPT-5.6 Sol and GPT-5.6 Luna as the supported models when billing through ChatGPT, with Fast mode unavailable on that route.
Models inside the apps v0 builds: this is separate from the model that writes your code. The AI models page explains that apps v0 generates connect by default to the Vercel AI Gateway, which v0 is pre-configured to use. Teams can also install marketplace integrations for fal, Deep Infra and Grok by xAI from Project settings, Integrations, or call a provider directly by adding its API key as an environment variable in Project settings.
Custom models: on team accounts, only a team owner can turn on "Allow custom models". With it off, the team is limited to v0 models.
Features and modes
The agentic features page and the 2025 and 2026 launch posts describe these capabilities.
| Feature | What it does |
|---|---|
| Chat to app | Each chat is a workspace for one app; follow-up messages edit it |
| Design Mode | Visual editing of the generated interface without prompting |
| Design systems | Design Systems 2.0 saves a design system as a skill; the API accepts up to three skills per request |
| Web search | Real-time search with inline source links |
| Browser use | Opens and uses the app it built, critiques the design, debugs flows and visits external URLs for reference |
| Automatic error fixing | Fixes missing files and dependencies, syntax issues, runtime errors and import or export problems |
| Fix with v0 | Button in the deployment popover that sends error logs to v0 for a fix |
| Terminal commands | Runs shell commands in the sandbox to test, inspect the repository, run unit tests and call the Vercel and GitHub CLIs |
| Permission modes | Ask, Auto and Full, controlling how freely the agent runs commands |
| Agent control | Stop at any time, auto-continue through multi-step tasks, each step shown in the chat |
| Marketplace integrations | Neon, Supabase, Upstash and Stripe among others, with tools v0 can call during generation |
| Data integrations | Connections to Snowflake and AWS databases for reports and data apps (February 2026) |
| MCP servers | Presets and custom servers (see below) |
| Git panel | Branch per chat, automatic commits, pull requests, deploy on merge |
MCP connections
From the + menu in the prompt form, MCPs offers preset servers for Contentful, Context7, Datadog, Glean, Granola, Hex, Linear, Mobbin, Nitrosend, Notion, PostHog, Sanity, Sentry, Slack, Vercel and Zapier. A custom server can use No Auth, Custom Headers, Bearer Token or OAuth 2.0 authentication.
Git workflow
The GitHub connection is optional and is set in Project menu, Settings, GitHub. Once connected, v0 creates a working branch, commits each change automatically and can open pull requests. The FAQ states that v0 never pushes directly to main. Publishing a GitHub-backed project merges the chat's pull request into the base branch and waits for the production deployment, and required checks and reviews still apply. Existing repositories can be imported if the connected GitHub account can access them.
Supported IDEs and platforms
| Surface | Details |
|---|---|
| Web app | v0.app in a browser; internet connection required |
| iOS app | Announced 24 November 2025, built with React Native and Expo; a chat experience for building on the go rather than a mobile copy of the website |
| v0 API v2 | Generally available 5 August 2026; base URL https://api.v0.dev/v2; synchronous, asynchronous (webhooks or polling) and streaming responses |
| SDK | The v0 package, plus @v0-sdk/ai-tools exposing v0 operations as tools for AI SDK agents, and a create-v0-sdk-app scaffold |
| v0 MCP server | https://v0.app/api/mcp, OAuth on first connection; tools to create and list chats, send messages and get preview URLs, usable from an IDE |
| Local editing | Export code to work locally, or sync both ways through the GitHub integration |
v0 generates web apps. The FAQ names Next.js, React, TypeScript, Tailwind CSS, shadcn/ui and the AI SDK (version 6 by default) as its strengths. Prompts can be written in any language and v0 replies in the same language. The full experience, with environment variables, deployments and previews, needs the repository connected to a Vercel project, but code can be exported and deployed elsewhere.
API inputs and previews: an API chat can start from a prompt, a GitHub repository, a ZIP archive or a set of files. Each chat gets a short-lived preview token, and Vercel recommends proxying preview requests through your own server route so the API key never reaches the browser. Chats created on the previous API version do not run on v2 and must be downloaded as ZIPs and recreated.
Enterprise and admin controls
v0 team administration sits on top of Vercel teams. The enterprise documentation describes these controls.
- SSO and Directory Sync: admins map an identity provider group to a Vercel Access Group, and members of that group receive the configured v0 access automatically. SAML SSO is listed on the Enterprise plan.
- Access Groups: one Access Group can set team roles, team permissions, project roles and v0 roles together. Synced members default to the Contributor team role unless a mapping assigns another.
- v0 roles: Builder, Creator and Viewer, separate from Vercel team roles such as Owner and Viewer. Integrations Manager is a separate team permission for managing integrations.
- Builder scope: a v0 Builder can create deployments and manage environment variables on every v0-connected project in the team, and the docs advise assigning it only where that breadth is intended.
- Custom model control: "Allow custom models" can only be turned on by a team owner; individual model restrictions still apply.
- Chat visibility: only people with access, everyone on the team (optionally view-only), anyone with the link, or anyone on the web.
- Restrict Chat Sharing: an owner setting that removes the "anyone with the link" and "anyone on the web" options.
- Owner access: Vercel team Owners can view, edit and archive any team chat opened by URL, including chats set to people with access only.
Availability: per the pricing page, centralised billing and team sharing start on Plus, the training opt-out is on by default on Business, and SAML SSO, role-based access control, priority access and support SLAs are listed on Enterprise.
Data handling and privacy
- Training: the FAQ states that v0 may use prompts and content as inputs to models and learning systems, including from third-party providers, to improve its products. Content from Enterprise customers and from customers who opt out in their Vercel team settings is not used. Customer data from Vercel's platform services is not used to train, improve or fine-tune v0's models.
- Opting out: team data preferences in Vercel settings control the training opt-out.
- Export and deletion: users can download their prompts and content. Deletion of v0 outputs can be requested from [email protected]; the FAQ notes that models already trained on inputs or outputs are not updated.
- Ownership: Vercel does not claim ownership of code generated from your prompts. The FAQ adds that output may resemble other users' output or third-party material and may be incomplete or buggy, so the customer is responsible for evaluating it.
- Attestation: Vercel holds a SOC 2 Type 2 attestation covering Security, Confidentiality and Availability, and v0 is within its scope.
- Chats: visibility settings decide who can read a chat; a chat set to "anyone on the web" is public.
The documentation pages fetched for this guide do not state a retention period for chats or prompts, so none is given here; check Vercel's privacy notice for current terms.
Release history
Dates from Vercel's blog posts, as published.
| Date | Change |
|---|---|
| 2024 | v0 becomes generally available (as stated in the February 2026 announcement) |
| 11 August 2025 | v0.dev renamed v0.app; agentic features: research, planning, web search with citations, reading files, inspecting live sites, task tracking |
| 24 November 2025 | v0 iOS app announced |
| 3 February 2026 | "The new v0": rebuilt on a sandbox runtime that imports any GitHub repository, Git panel with branch per chat and pull requests, Snowflake and AWS database connections |
| 5 August 2026 | New v0 API generally available: headless agent, Vercel Sandbox previews, streaming, SDK, v0 MCP server |
Limits and known constraints
- Free plan cap: 7 messages per day.
- Credit consumption varies by model: v0 Max Fast output costs 50 times v0 Mini output per token, so model choice drives spend.
- Web apps on a set stack: v0 is strongest on Next.js, React, Tailwind CSS and shadcn/ui; it builds web apps rather than native desktop software.
- Online only: v0 cannot be used offline.
- Vercel coupling: environment variables, previews and deployments need a Vercel project; export works for other hosts.
- API migration: chats from the previous API version must be recreated on v2.
- ChatGPT route: Fast mode is unavailable and credits are still needed for platform fees.
- Rate limits: the API has documented rate limits; see the API reference.
Security guidance
v0 sits on a shared responsibility line: Vercel runs the platform, and the customer decides who can build, which data the agent can reach and what gets deployed. The settings below are customer-side.
- Scope roles tightly: give v0 Builder only to people who should deploy and manage environment variables across every v0-connected project. Use Creator or Viewer otherwise.
- Turn on Restrict Chat Sharing so chats cannot be made link-public or web-public, and review existing chat visibility.
- Treat environment variables and MCP credentials as secrets: prefer OAuth 2.0 for custom MCP servers where offered, use scoped tokens, and rotate keys added for provider APIs. See the Vercel Tokens Security Guide.
- Keep the API key server-side: proxy preview requests through your own route as v0's API guidance describes.
- Use Ask permission mode for terminal commands on sensitive repositories, and keep branch protection and required reviews on the base branch so v0's pull requests go through normal review. See the AI-Generated Code Review Guide.
- Set the training preference deliberately in team settings before importing proprietary repositories.
- Govern MCP servers: allow only the preset or custom servers the team has assessed. See the MCP Security Guide.
For the wider practice of securing prompt-built apps, see the Vibe Coding Security Guide.
Practitioner checklist
- Confirm which plan each team is on and whether anyone is still on the sunsetting Premium plan.
- Set the team training opt-out (or confirm Business or Enterprise defaults) before importing private code.
- Map identity provider groups to Access Groups and review who holds the v0 Builder role.
- Turn on Restrict Chat Sharing and audit chats set to link or web visibility.
- Decide whether "Allow custom models" should be on, and record who approved it.
- Connect GitHub with branch protection and required checks on the base branch.
- List every MCP server and marketplace integration in use, with its authentication method and owner.
- Set credit spending expectations per model and review v0.app/settings/usage monthly.
- For API users, keep keys server-side and plan migration of any v1 chats to v2.
Standards and references
- v0: Documentation
- v0: Pricing
- v0: FAQs
- v0: Enterprise
- v0: Agentic features
- v0: MCP integrations
- Vercel: Introducing the new v0
- Vercel: Introducing the new v0 API
Related NHI Mgmt Group resources: Vercel Tokens Security Guide · Vibe Coding Security Guide · Vibe Coding Explained · AI-Generated Code Review Guide