Bolt.new

Bolt.new takes a slightly different route to AI app building.

You describe what you want, and the platform generates a working application that you can view and continue editing directly in your browser. But unlike simple AI website generators that mainly produce a front-end mockup, Bolt is built around a real development environment.

That distinction comes from StackBlitz’s WebContainers technology. Bolt can run a Node.js development environment inside the browser, giving the project access to a code editor, terminal, packages, development server, and live preview without requiring a traditional local setup.

The result is a product that sits somewhere between an AI app generator and a browser-based development environment.

That makes Bolt.new particularly interesting for people who want AI to handle the initial implementation but still want to see and work with the actual project underneath.

Bolt.new Starts With a Conversation, But It Does Not Stop at a Mockup

The basic workflow is simple.

You describe the application you want in natural language. Bolt interprets the request, creates the project structure, installs the required packages, writes the code, and launches the result inside the browser.

From there, you can continue the conversation.

Instead of manually opening multiple files to make a small change, you can describe the next adjustment. You might ask for a different navigation layout, a new form, an authentication screen, a database connection, or another page.

Bolt then modifies the project.

This conversational workflow is one of the reasons the platform fits the growing prompt-to-app category. Current independent comparisons consistently describe Bolt as a tool for generating working applications from natural-language instructions, particularly for prototypes and early-stage products.

But the important difference is that the generated project remains a real codebase.

Bolt’s own product documentation says it creates complete applications, including front-end and back-end components, and gives users access to the actual code rather than keeping the result inside a closed visual editor.

That makes the product more interesting for developers than a conventional no-code builder.

The Browser Is Actually Part of the Product

The most distinctive technical idea behind Bolt is not the chat box.

It is WebContainers.

StackBlitz developed WebContainers to run Node.js environments directly inside a browser. Instead of sending every development task to a conventional remote development server, the environment runs within the browser using WebAssembly and browser technologies.

For Bolt users, this means the application can start quickly and behave much more like a development environment than a typical website builder.

You can work with files, dependencies, a terminal, a development server, and a live application preview without first installing Node.js and configuring a local project.

That is especially useful when the goal is experimentation.

Suppose you want to test a new React package, build a dashboard concept, reproduce a bug, or create a small SaaS interface. The normal process might involve creating a project, installing dependencies, configuring the environment, and then starting development.

Bolt compresses much of that setup into the browser.

The browser-first architecture is therefore more than a convenience feature. It explains why Bolt feels different from AI assistants that sit inside an existing desktop IDE.

You Still Get the Code Behind the AI

One concern with AI-generated applications is what happens after the initial demo.

A pretty interface is easy to generate. Maintaining the project later is a different problem.

Bolt’s approach gives users access to the underlying files and code. That matters because the generated application can be inspected, modified, connected to external services, and taken into a broader development workflow.

GitHub integration also gives teams a path toward conventional source control.

This is an important distinction when comparing Bolt with more restrictive no-code platforms. The goal is not simply to create something that looks like an application. The platform is designed to create a project that developers can continue working with.

That does not mean every generated codebase will automatically be clean or production-ready.

AI-generated projects still need human review. Dependencies can conflict, application logic can be incomplete, and a generated authentication or payment implementation should never be accepted without appropriate security review.

Bolt can reduce the amount of initial development work. It does not remove the need to understand what the application is doing.

Bolt Cloud Brings More of the Launch Process Into One Place

Bolt has expanded beyond application generation with Bolt Cloud.

The current platform can provide hosting and database functionality from the same project environment. Bolt’s documentation says Bolt Cloud can automatically create databases, provide database management features, handle authentication and users, manage secrets and edge functions, and host projects.

Every project can receive a Bolt-hosted subdomain, while paid plans add custom-domain support and additional publishing options.

That makes the workflow relatively compact:

Describe the product.

Generate the application.

Connect the data layer.

Test the project.

Publish it.

The alternative is to create the application in one service, configure a database somewhere else, find a hosting provider, connect the domain, and manage several accounts.

Bolt does not eliminate those external services, though.

The platform supports integrations with services such as Supabase, Netlify, GitHub, Stripe, and Figma. This is useful because a project does not necessarily have to remain entirely inside Bolt’s own infrastructure.

The company specifically presents Bolt Cloud as an option rather than a mandatory replacement for external services.

The Technology Stack Has a Clear Boundary

Bolt is flexible, but it is not technology-neutral.

The platform focuses heavily on JavaScript-based web development. Official documentation says Bolt supports Node.js backends and browser-based JavaScript frameworks, while Python and PHP backends are not supported in the standard Bolt environment.

That distinction is easy to overlook when an AI builder promises to create applications from natural-language instructions.

The AI can generate a lot of code, but the underlying platform still determines what kinds of projects are practical.

For React, Vue, Svelte, Next.js, Nuxt, Astro, and other JavaScript-oriented web projects, Bolt can be a natural fit.

For a backend that specifically depends on Python, PHP, or another unsupported server-side environment, the workflow becomes less straightforward.

This is one area where Bolt’s speed should be considered alongside its technical boundaries.

Mobile Apps Are Possible, But the Workflow Is Different

Bolt is primarily a web development environment, but it can also be used for mobile application projects through Expo.

The official documentation says users can direct Bolt to create an Expo-compatible application by including “mobile app” in the initial prompt.

That gives the platform a route into mobile development without turning Bolt into a native mobile IDE.

This distinction matters.

A web application and a mobile application may share concepts, but they have different build requirements, device APIs, testing processes, packaging requirements, and publishing workflows.

Bolt can help create the project, but teams building a serious mobile product should still understand the Expo and React Native ecosystem that sits underneath it.

For quick mobile prototypes and early validation, the feature can be useful. For a complex native product, additional development work should be expected.

Pricing Is Really a Token Conversation

Bolt.new has a free plan, but its pricing is not best understood as simply “free versus $25.”

The important variable is token usage.

The current Free plan costs $0 and includes a 300,000-token daily limit and 1 million tokens per month. It also includes public and private projects, website hosting, unlimited databases, and Bolt branding.

The Pro plan costs $25 per month when billed monthly. It starts with 10 million tokens per month and removes the daily token limit. It also adds features such as custom domains, private site sharing, higher file-upload limits, SEO boosting, expanded database capacity, and the ability to choose a database provider.

Teams costs $30 per member per month on monthly billing and adds centralized billing, team access controls, organization sharing, private NPM registry support, and design-system knowledge.

Enterprise pricing is custom.

The important point is that Bolt measures AI usage in tokens. Larger projects can consume more tokens because Bolt needs to process and synchronize more of the project when responding to requests.

That means the effective cost of building something can depend on the size and complexity of the project, not only on the subscription price.

Paid tokens can roll over for one additional month, provided the subscription remains active.

Where Bolt.new Makes the Most Sense

Bolt is particularly well suited to projects where speed matters.

A founder might want to turn a product concept into an interactive MVP.

A designer might want to convert an interface idea into a working prototype.

A developer might want to test a framework or quickly create a functional proof of concept.

A small business might need a custom internal tool without starting a traditional development project from zero.

These situations share one characteristic: the first working version is valuable.

Bolt’s browser environment removes much of the setup friction, while the AI handles a large amount of the initial implementation.

The product becomes less attractive when the project requires a technology stack outside its supported environment or when the codebase grows into a highly specialized engineering system that needs extensive manual architecture work.

That does not mean Bolt cannot be used for serious applications. It means the amount of human engineering required can increase substantially as the project moves beyond its original generated structure.

Bolt’s Real Competition Depends on What You Need Next

Bolt is often compared with Lovable, Replit and v0, but these products approach AI development differently.

Lovable is closely aligned with the conversational full-stack app-builder model.

Replit combines AI development with a broader cloud development environment.

v0 is strongly associated with UI and React/Next.js generation within the Vercel ecosystem.

Cursor takes another route entirely by focusing on AI-assisted development inside a code editor.

Current comparison coverage repeatedly identifies Bolt’s browser-based environment and rapid prompt-to-working-app workflow as key differentiators, while token usage and long-term iteration remain common discussion points.

So the right comparison is not simply “which AI builder is better?”

The more useful question is what you want the tool to do after the first version appears.

If the priority is turning an idea into a working browser-based application quickly, Bolt’s architecture makes sense.

If the priority is long-term software engineering inside a traditional repository and development environment, a coding agent or IDE-centered tool may fit the workflow more naturally.

The Best Way to Think About Bolt.new

Bolt.new is easiest to understand as an AI application builder sitting on top of a genuine browser development environment.

That combination is what makes it different.

The AI handles much of the initial implementation.

WebContainers provide the development environment.

Bolt Cloud can handle hosting and databases.

External integrations provide additional infrastructure.

The underlying code remains visible and usable.

The biggest practical constraint is the same feature that makes the platform easy to start with: the AI has to process the project as it evolves, and larger projects can consume more tokens.

That makes project size, iteration style, and technology choice important when evaluating the platform.

For prototypes, MVPs, web applications, internal tools, and fast product experiments, Bolt.new offers a particularly direct path from an idea to something running in a browser.

For projects with unusual backend requirements, large existing codebases, or highly specialized production engineering needs, the browser builder model may not be enough by itself.

The platform’s value is therefore less about replacing every part of software development and more about shortening the distance between “I have an idea” and “I can open and use the first version.”