Bolt.new

Bolt.new is an AI-powered full-stack app builder made by StackBlitz, launched on October 3, 2024. Users describe what they want to build in plain language, and Bolt generates, runs, and deploys a complete web application directly inside a browser tab. The core technical differentiator is WebContainers, StackBlitz's WebAssembly port of Node.js that runs the entire development environment client-side, meaning no Docker setup, no cloud VM spin-up, and no local install. StackBlitz CEO Eric Simons had been building toward this infrastructure since the company's founding in 2017; bolt.new was the moment that seven-year bet found its product-market fit.

Out of the box, Bolt handles the full stack: it scaffolds React, Next.js, Vue, Svelte, Astro, Remix, and Express projects; installs npm packages; runs a live dev server; wires up Stripe payments, Supabase databases, GitHub auth, and Figma imports; and publishes to a free bolt.host subdomain or a custom domain on paid plans. Mobile apps are supported via React Native with Expo. An open-source fork, bolt.diy, is maintained under MIT license and supports 19+ LLM providers for teams who want to self-host. The AI backend defaults to Anthropic's Claude models, with users able to switch models inside the interface.

What Bolt.new actually does in April 2026

A Bolt session works like a conversation with a developer who can see and run your code in real time. You type "build a SaaS dashboard where users upload CSV files and see charts," and within minutes the tool has scaffolded a React frontend, a Node.js API route, a file-upload handler, and a charting library, all running live in the same browser tab. You can then refine by prompting ("add a dark mode toggle" or "wire this to Supabase auth"), and Bolt updates the running app immediately.

In August 2025, StackBlitz launched Bolt Cloud, which extended bolt.new from code generation into full application infrastructure: managed databases, production-grade hosting at bolt.host, and one-click custom domain deployment without requiring external Netlify or Vercel accounts. Paid plans unlock private projects, AI image editing, and token rollover (unused tokens carry forward one month, effective July 2025). The free tier provides 1 million tokens per month with no local setup required.

"the dev environment is lightning fast, the deployment process is effortless" -- Gregory Walker, Product Hunt, October 2024

Where Bolt.new sits versus Lovable and v0

Three tools dominate the "natural language to working app" space in 2026: Bolt.new, Lovable, and v0 by Vercel. They solve related but distinct problems, and the mechanical differences matter more than marketing language suggests.

Bolt.new vs. Lovable: Both are browser-based full-stack builders with similar prompt-to-app workflows. The architectural split is the database layer. Lovable has deep Supabase integration baked in: it automatically provisions database tables, writes row-level security policies, and wires up authentication without the user touching any configuration. Bolt.new requires users to manually connect Supabase (or use Bolt Cloud's managed database, which launched August 2025). Lovable runs its builds on cloud VMs rather than WebAssembly, meaning cold starts are slower but it handles heavier server-side workloads more reliably. Lovable uses a credits-based pricing model; bolt.new uses tokens, which behave differently under debugging-heavy usage. For Supabase-first backends, Lovable has the more seamless path; for speed and the ability to prototype without any accounts, Bolt wins.

Bolt.new vs. v0 (Vercel): v0 is a component-first tool with a completely different scope. It generates clean, production-quality React components using Shadcn, Radix, and Tailwind, designed to drop into an existing codebase. There is no backend, no database, no running dev server, and no deployment. v0 outputs code for you to copy into your own project. Bolt generates entire apps end-to-end and runs them live. v0's free tier is more generous for pure UI work because component generation consumes fewer resources than full app execution. The practical pattern many developers use: v0 for component-level UI refinement, Bolt for scaffolding the running app, then export to a local IDE for production hardening.

What the daily build loop actually looks like

For straightforward projects, bolt.new delivers on its promise. Landing pages succeed at roughly a 92% rate and take 2-4 minutes. Basic dashboards with authentication come in around 68% success rate and 25-35 minutes. E-commerce MVPs drop to about 52% (45-60 minutes). Complex SaaS applications with custom business logic fall to roughly 31% and 90+ minutes, based on hands-on testing documented in 2025.

The debugging experience is where the economics get uncomfortable. Bolt uses a conversational fix-it loop: when something breaks, you describe the problem and Bolt generates a patch. The problem is that each exchange, including the AI reading your entire project context, burns tokens. A single authentication bug that takes three attempts to fix can consume 3-5 million tokens alone. Users on the Pro plan (10 million tokens per month) report burning through their monthly allocation in a single bad debugging session.

"when you request changes, it often makes unintended adjustments to other components" -- Rajat Sharma, Product Hunt, 2024

Bolt does offer a "diffs" mode that prevents full-file rewrites for small changes, potentially saving millions of tokens per session. The problem is it ships turned off by default. Many users, especially those new to the platform, do not discover it until after their first costly session.

Who Bolt.new is built for

Bolt's clearest use case is the non-technical founder or indie hacker who needs a working prototype, not production code. The $3,700/month SaaS product built in 8 weeks that one Reddit user described (versus a contractor estimate of 8 months and $19,000) is a real use case, not an anomaly. For simple tools, internal dashboards, landing pages, and MVPs where correctness is more important than code quality, bolt.new dramatically compresses the time from idea to something deployable.

Developers use it for rapid scaffolding: generate the boilerplate, validate the idea, export the code to GitHub, then continue in Cursor or a local IDE. The WebContainers approach means the generated environment matches what you'd get locally, making code export cleaner than tools that run on remote VMs with proprietary configurations.

The fastest 0-to-40M-ARR sprint in startup history (the $40M ARR milestone was reached by March 2025, five months post-launch) happened because bolt.new found a population of people who were previously locked out of building software entirely. Those users do not care about code quality or scalability. They care about having something that works and can be shown to a customer.

What Bolt.new is not

Bolt is not a path to production-grade software for most complex applications. The success rate on complex SaaS apps (31%) reflects a genuine ceiling, not just a skill issue. When projects need custom state management, real-time features, security-sensitive logic (payment processing, auth flows with edge cases), or non-JavaScript backends (Python, Go, Ruby, PHP), bolt.new is not the right tool.

It is also not a tool with predictable costs for power users. The token economy means a complicated week of debugging can cost more than the monthly subscription suggests. The Trustpilot rating of 1.4 out of 5 (from 140+ reviews as of 2025) is dominated by billing confusion and support complaints, not product failures. Users who need cost certainty should budget conservatively or use the free tier for experimentation before committing to paid plans.

Teams expecting collaborative workflows will find it limiting. There is no version control, no branch-based development, no code review process, and no rollback in the hosted flow. Engineers working together on a codebase still need a real development environment alongside or instead of Bolt.

Finally, while bolt.diy is open source (MIT license), bolt.new itself does not expose a public API for programmatic access. You cannot call bolt.new from your own scripts or applications; it is a product you use through its web interface.