Bolt.new

Bolt.new is a browser-based AI app builder from StackBlitz that turns natural language prompts into working full-stack applications. Powered by WebContainers, a WebAssembly-based micro OS that runs Node.js entirely inside your browser tab, Bolt generates code, installs dependencies, and gives you a live preview without touching a server. You can iterate by chatting, deploy to Netlify or Cloudflare in a single click, and export the full source code whenever you're ready to take over. Launched in October 2024, it became the reference product for the prompt-to-app category almost overnight. The free tier is genuinely usable. The Pro plan at $25/month is reasonable for frequent builders. Experienced developers use it for fast prototyping. Non-developers use it to ship things they never thought they could build.

In October 2024, StackBlitz shipped Bolt.new and the conversation about AI coding tools changed. Before Bolt, you needed a local dev environment, a cloud IDE, or at minimum a working knowledge of npm to get something running. Bolt changed that equation: type a description into a chat box, wait a few seconds, and you have a running full-stack application inside your browser tab. No setup. No install. No waiting for a VM to boot. The viral moment that followed wasn’t hype for its own sake. It was people watching something that genuinely hadn’t existed before.

Quick verdict

Bolt.new is the best starting point for anyone who wants to go from idea to deployed app in under an hour. For experienced developers it’s an excellent rapid prototyping tool that cuts the boilerplate tax significantly. For non-developers it’s closer to magic than anything else currently available. The WebContainers technology underneath is more interesting than most people realize, and the free tier is real enough to make it worth trying before you spend anything.

What is Bolt.new, exactly?

Bolt.new is an AI app builder that runs inside a browser tab. You describe what you want, the AI generates the code, and a live preview appears alongside the editor without you touching a terminal or a package manager. That’s the pitch. What makes it technically different from every other tool in the space is the engine underneath it: WebContainers.

StackBlitz has been building browser-based development tools since 2017. For most of that time they were known as the place you’d send someone to open a code example without installing anything. In 2021 they published WebContainers, a WebAssembly-based micro operating system that boots a full Node.js environment inside a browser tab. Not simulated. Not stripped down. A real Node.js runtime, running locally, with npm install support and a working dev server.

The difference from legacy cloud IDEs is structural. Traditional online IDEs stream output from a remote server. That means cold starts, network latency, and someone else’s compute bill. WebContainers move all of that into your browser. The dev server runs on localhost inside your tab, so boot time is milliseconds and the live preview has zero network latency.

Teams at Google, Meta, Shopify, and Stripe were running WebContainers before Bolt existed. The viral product is new. The infrastructure underneath it is not.

The features that drove the viral moment

Prompt to running app in seconds

The core experience is simple enough that it takes thirty seconds to describe and two or three minutes to actually try. You open bolt.new, type something like “build a to-do app with a dark theme and local storage persistence,” and the AI starts generating. Within seconds you have a file tree, a code editor, and a live preview running side by side. The app is actually running. You can click buttons, add items, and see state persist. It’s not a screenshot of code. It’s a working application.

What improved over time is the quality of that first generation. Early versions would break on complex prompts and require several fix cycles. By 2025, Bolt claimed a 98% error reduction through automatic testing and refactoring. That specific number is hard to verify across project types, but the qualitative improvement in first-pass usability is real.

WebContainers, Node.js in your browser

This deserves more than a bullet point because it’s why Bolt can offer the free tier it does and still deliver a fast experience. Every time you start a new Bolt project, a Node.js environment boots inside your browser tab using WebAssembly. The npm install that would take a minute on a cold cloud VM takes seconds locally. The dev server that would have a 200ms round-trip latency because of network hops has zero latency because it’s on localhost.

The practical implication is that the live preview in Bolt doesn’t lag. When you ask it to change the button color, the change shows up in the preview almost immediately, because the dev server watching for changes is running inside your tab, not in a data center.

The constraint is real but bounded: WebContainers can run anything that compiles to WebAssembly or runs in Node.js. Native binaries and system-level packages are out. For most web applications, that constraint never surfaces. React, Vue, Node APIs, and SQLite-backed apps all run fine.

Iterate by talking, not by editing

After the initial generation, you keep building by continuing the conversation. “Make the sidebar collapsible.” “Add user authentication with email and password.” “The submit button isn’t working on mobile, fix it.” Bolt takes each message, reads the current state of the codebase, and applies targeted changes. You don’t have to explain where things live or which file to edit. The AI holds the context of what it built.

This is where non-developers find their footing. A product manager can specify what they want with product language and get changes applied correctly most of the time. The failure modes are real: deep architectural changes and subtle state management can require several rounds of back-and-forth to resolve. But the hit rate on straightforward UI and logic changes is high enough to keep non-developers in flow rather than getting stuck.

Deploy to Netlify or Cloudflare in clicks

Once you have something worth showing, deployment is two or three clicks. Bolt has native integrations with Netlify and Cloudflare Pages. You authorize the connection once, and after that you can push a live URL from inside the Bolt interface without leaving the browser. For a prototype you’re showing to stakeholders or investors, this is a significant time saver. The deploy pipeline that would normally require configuring a CI/CD workflow and managing environment variables is handled automatically.

Custom domain support is available on the Pro plan. On the free tier, your deployed site carries Bolt branding. For most prototyping use cases, that’s fine. For client work or anything you’re showing publicly as your own product, the Pro plan upgrade makes sense on that point alone.

Export the code when you outgrow it

You’re never locked in. Any Bolt project can be exported as a zip file or pushed directly to a GitHub repository. The export is the actual source code that Bolt generated. When you open it locally, it behaves the same way it did in Bolt, because the same Node.js app that ran in the WebContainer runs fine in a standard development environment. The handoff to a normal engineering workflow is not painful. This matters because the honest use case for Bolt is often “get to a working prototype as fast as possible, then hand it to an engineer or continue it yourself with better tools.” That path works.

Pricing

Bolt.new uses a token-based freemium model. Four tiers as of May 2026.

Free is $0, with 1 million tokens per month and a 300,000 daily cap. That daily cap is the real constraint. A complex build session can burn 50,000 to 150,000 tokens, so a serious full-day session will hit it. For exploration or building one thing at a time, it’s plenty.

Pro is $25/month with 10 million tokens, no daily cap, token rollover, 100MB file uploads, custom domain support, and no Bolt branding on your sites. For anyone using Bolt more than once a week, the math favors upgrading. It was $20/month at launch and moved to $25 in 2025, so older reviews may show the wrong price.

Teams is $30/user/month and adds centralized billing, admin controls, private NPM registry support, and design system integrations. It’s aimed at product agencies and small teams who want a shared workspace.

Enterprise is custom, with SSO, audit logs, compliance support, and dedicated account management.

One genuine frustration: token costs per request aren’t shown before you send. You can end a session unsure how much headroom you have left. A usage dashboard exists, but token accounting in this category still needs work.

Where Bolt.new wins and where it doesn’t

Bolt wins on speed to first working thing. Nothing gets you from blank page to deployed app faster, and nothing that gets you there is as accessible to non-technical users. The browser-native architecture means you don’t need to configure anything. The free tier means you can evaluate it with real projects before paying. The export means the investment of time building something in Bolt isn’t stranded if you want to move it.

Bolt wins on front-end web apps. React, Vue, Svelte, landing pages, dashboards, simple CRUD apps with a SQLite or Postgres backend: this is where the tool is reliable and fast.

Bolt struggles with complexity at scale. Projects with more than a handful of interconnected services, complex state management patterns, or requirements that depend on architectural decisions the AI makes early on can produce apps that work but are built in ways that cause problems later. When you export and open the code, you sometimes find patterns you wouldn’t have chosen. That’s fine for a prototype. It’s a problem if you’re planning to hand this to an engineering team as a production starting point without a cleanup pass.

Bolt also struggles when your requirements conflict with WebContainers constraints. If your project needs a language or runtime that doesn’t run in the browser, you’re blocked in a way that no amount of iteration will fix.

Who Bolt.new is built for

The product serves two distinct groups well and a third group in a more limited way.

Founders and product people who need to validate an idea fast are the primary audience. If you can describe your product, you can build enough of it in Bolt to show it to users or investors. That value is real and significant. The barrier between “I have this idea” and “here’s a link you can click” used to be measured in weeks. Bolt makes it hours.

Developers who do a lot of prototyping are the second group. If you’re a freelancer or consultant who regularly needs to stub out projects, Bolt removes the boilerplate-heavy part of the work and gets you to the real engineering faster.

Non-technical users who want to build internal tools or simple products are the third group. Bolt works well for them on straightforward requirements. When the app needs something unusual, the gap between what they can describe and what the AI reliably delivers starts to show.

Bolt.new vs the alternatives

v0 from Vercel is the most common comparison but the scope is different. v0 generates polished React UI components you drop into an existing Next.js codebase. It’s not trying to build a full application. Bolt is. If you’re starting from zero and want a deployed app, Bolt is faster. If you already have a codebase and need a component built to Vercel’s conventions, v0 fits better. Full breakdown at Bolt vs v0.

Lovable is the closest direct competitor. It also targets non-technical users, generates full-stack apps, and connects to Supabase. Lovable tends to produce more polished UI from the first prompt. Bolt has more developer-facing flexibility and the WebContainers architecture gives it a structural performance edge. The choice usually comes down to which interface you prefer. See Lovable vs Bolt.

Replit Agent runs in a cloud VM, not the browser. That means it can run Python, arbitrary packages, and anything outside the Node.js ecosystem. If your project needs a Python backend or ML components, Replit Agent is the right tool. For pure web apps, Bolt is faster and cheaper to start.

For a broader view, the best AI agent for coding guide covers the full landscape.

Getting started

Go to bolt.new. You don’t need an account to try it; you can start generating with a prompt immediately, though you’ll need to sign up to save your project. Use a GitHub account for the fastest onboarding.

Your first prompt should be specific. “Build a to-do app” will produce something generic. “Build a to-do app with a dark sidebar, tags for each item, and a filter to show only incomplete tasks” will produce something closer to what you actually want. More detail in the first prompt means fewer iteration cycles, which means fewer tokens consumed.

Once the app is generated, treat the chat as your editor. Describe what you want changed, not how to change it. Paste error messages directly into the chat when things break. Keep your browser tab open during a session, because WebContainers state lives in the tab and an unexpected refresh can lose unsaved progress. Connect a GitHub repository early if version history matters to you.

The bottom line

Bolt.new earned its viral moment. The combination of a credible AI code generator with WebContainers technology in an accessible browser interface is genuinely novel, and the execution is good enough that people ship real things with it. It’s not a toy, even if the free tier makes it feel like one to try.

The ceiling is real: complex production applications will outgrow it, and the AI’s reliability drops as architectural complexity increases. The right way to use Bolt is as a fast path to a working thing, not as a permanent home for serious engineering work. For founders, prototypers, and anyone who’s been blocked by the setup friction of modern web development, that fast path is exactly what was missing.

Key features

• Prompt-to-app: generate full-stack apps from natural language
• WebContainers run Node.js entirely inside the browser with no server required
• In-browser terminal, file editor, and live preview in a single tab
• One-click deployment to Netlify or Cloudflare Pages
• Export full project source code at any time
• Iterate by chatting with natural language instructions
• Unlimited databases included on all plans

Pros and cons

Who is Bolt.new for?

• Founders and product managers who need a working prototype before they can justify engineering time
• Freelance developers who want to stub out client projects in an afternoon rather than a week
• Marketers and designers who can describe exactly what they want but can't write the code themselves
• Students learning web development who want to see a real app running before they understand all the pieces

Alternatives to Bolt.new

If Bolt.new isn't quite the right fit, the closest alternatives are v0 , lovable , and replit-agent . See our full Bolt.new alternatives page for side-by-side comparisons.