---
title: "Bolt.new review: pricing and what it is really worth"
description: "Is Bolt.new worth $25 a month? Verified prices, how tokens really work, strengths, limits and 6 settings that make your tokens last longer."
url: "https://vibe-code.blog/en/bolt-new-review"
lang: "en"
author: "Thierry Gustin"
published: "2026-09-08"
updated: "2026-11-21"
---

# Bolt.new review: pricing and what it is really worth

> **Key takeaways** — Bolt.new is a StackBlitz app builder that runs your whole project in the browser, with hosting, a database and domains built into Bolt Cloud. The free plan gives 1 million tokens a month (300K a day) and Pro starts at $25 a month for 10 million tokens. It is an excellent prototyping tool whose cost rises with project size.

Bolt.new is a browser-based app builder made by StackBlitz: you describe your site or app, Bolt writes the code, runs it right in your browser and can publish it with a database and a domain name. Our Bolt.new review in short: it is one of the fastest ways to get a working prototype while keeping real access to the code, but its token use climbs as the project grows. Here are verified prices, how tokens actually work and what Bolt is worth for your kind of project.

This review draws on Bolt's official documentation, pricing page and release notes. To see how it stacks up against Lovable, Replit or Cursor, check our [comparison of the best vibe coding tools](https://vibe-code.blog/en/best-vibe-coding-tools).

## How does Bolt.new work?

Bolt.new runs your entire project in the browser thanks to WebContainers, the StackBlitz technology that runs Node.js without a remote server. You install nothing, and the code, the terminal and the preview all sit in one place.

According to the help center, a Bolt project includes:

- **A chat with an agent** that builds and edits the app. There are two agents: Standard, "balanced for everyday building," and Max, designed for complex tasks and large projects. Bolt picks the underlying models behind the scenes.
- **Plan mode**, which replaced the old Discussion mode, for thinking before the code changes.
- **A Code view** where you can edit files by hand without spending tokens.
- **Bolt Cloud**: hosting with a free bolt.host subdomain, a database created automatically when the app needs one, and domain management.
- **Imports**: Figma designs, GitHub repositories or Lovable projects can be your starting point.
- **Mobile apps** go through the Expo integration.

Bolt builds web applications with "a wide range of JavaScript-based web frameworks," per the docs.

## How much does Bolt.new cost?

Bolt.new has a free plan, Pro from $25/month, Teams at $30 per member per month and custom-priced Enterprise. Prices checked on 21 November 2026 on [Bolt's official pricing page](https://bolt.new/pricing), in US dollars before tax.

| Plan | Price | Tokens | What you get |
| --- | --- | --- | --- |
| Free | $0 | 1M a month, 300K a day | Hosting, databases, Bolt badge on sites, uploads up to 10 MB |
| Pro | From $25/month | From 10M a month | No daily limit, custom domain, no badge, uploads up to 100 MB, token rollover |
| Teams | $30 per member per month | Depends on plan | Centralized billing, access management, admin controls |
| Enterprise | Custom | Custom | SSO, audit logs, priority support, SLA |

A few useful details:

- **Yearly billing** saves "up to 28%," according to the pricing page. The help center gives the example of Pro 10M at $216 a year, which works out at $18 a month.
- **Higher Pro tiers** exist: the help center mentions, for example, a $100-a-month tier with 55 million tokens.
- **Rollover**: on paid plans, unused monthly tokens stay valid for two months, and the oldest are used first. If you cancel, you lose all your tokens at the end of the billing cycle.
- **The free plan** rolls nothing over: daily tokens reset every day and monthly tokens on the 1st.

## What is a Bolt token, and why does the bill climb?

A token is a small piece of text the AI reads or writes. The key point, which Bolt's own help center spells out: most token use comes from reading, understanding and syncing your project's files. The bigger your project, the more each message costs.

That is the big difference from tools that charge per message. On Bolt, the tenth message in a small project is cheap; the hundredth message in an 80-file project costs noticeably more, even for a tiny change.

> **The big-project trap** — A prototype that keeps growing without any cleanup costs more tokens per message. Delete unused files, split big requests and make small tweaks yourself in Code view.

### 6 settings that make your tokens last

1. **Make small changes by hand** in Code view: a label, a color, a margin. It uses zero tokens.
2. **Use visual edits** in the preview: the help center says they are saved as one batch and only use tokens when you save.
3. **Go through Plan mode** for decisions: an approved plan saves you rebuilding the same page three times.
4. **Save the Max agent** for complex work (refactors, interconnected features) and use Standard for the rest.
5. **Write precise requests**: one request per feature, with the expected result. Our [5-step prompting method](https://vibe-code.blog/en/vibe-coding-prompts) helps.
6. **Create skills**: Bolt lets you save reusable instructions that apply automatically when a prompt matches, on every plan. You stop repeating your conventions in each message.

## What Bolt.new does well

Bolt.new is very good at getting you from an idea to something that runs, while keeping the code in plain view.

- **Fast start**: nothing to install, the app runs in the browser from the first message.
- **Direct code access**: Code view and the terminal are right there, which makes Bolt a good place to learn how an app is put together.
- **Built-in infrastructure**: hosting, a database with email or Google sign-in, file storage and domains, all in Bolt Cloud. On Pro or Teams, you can use your own Supabase database instead.
- **Imports**: starting from a Figma design or an existing GitHub project saves a lot of time.
- **Mobile**: the Expo integration opens the door to iOS and Android apps. See our guide to [building a mobile app with vibe coding](https://vibe-code.blog/en/mobile-app-vibe-coding).
- **Security audit**: on paid plans, a project-wide audit is available from the Publish menu and does not use tokens.
- **Code ownership**: the docs state that you own your code and can use it for any legal purpose.

## Bolt.new's limits

Bolt.new has clear weaknesses that you should know before building a product on it.

### Cost depends on project size

As covered above, every message rereads part of the project. Large projects get expensive to keep changing, which often pushes people to move the code to a coding agent. Our breakdown of [what a vibe-coded app really costs](https://vibe-code.blog/en/vibe-coding-cost) lists every line item.

### Hard limits on the free plan

Sites published on the free plan are capped at 10 GB of bandwidth and 333,333 requests a month, with a hard limit. The help center is explicit: once a free site hits the limit, it stops serving content until the next cycle starts. On Pro, the cap rises to 30 GB and 1 million requests, with pay-as-you-go capacity beyond that.

### JavaScript only

Bolt builds apps with JavaScript frameworks. For a Python script, a scheduled server task or data processing, [Replit](https://vibe-code.blog/en/replit-agent-review) is a better fit.

### Security is still on you

A built-in database does not guarantee that each user only sees their own data. Always test with two accounts and work through our [pre-launch security checklist](https://vibe-code.blog/en/vibe-coding-security). Access rules are covered in our article on [Supabase RLS](https://vibe-code.blog/en/supabase-rls-security).

## A starter prompt for your first Bolt project

This prompt keeps token use down by framing the project from the first message. Start it in Plan mode.

```text
Project: [type of app] for [users].
Version 1, nothing else:
1. [feature]
2. [feature]
3. [feature]
Constraints:
- As few files as possible, no unnecessary dependencies.
- Simple, mobile-friendly interface, in English.
- If a database is needed, each user only sees their own data.
Start with a numbered plan listing the files you will create. Wait for my approval before writing code.
```

Once the plan is approved, build one feature at a time and test it before moving on. Connect GitHub from the start: our guide to [Git for vibe coders](https://vibe-code.blog/en/git-for-vibe-coders) explains why.

## Who is Bolt.new for?

Bolt.new is best for prototypes, websites and small web apps, and for people who want to see the code.

- **Yes, if** you want to test an idea in one evening, show a clickable prototype to a client or learn how a web app is built.
- **Yes, if** you want a test mobile app through Expo.
- **Consider Lovable** if you want an app with user accounts and as little technical work as possible: see our [Lovable review](https://vibe-code.blog/en/lovable-review).
- **Consider Replit** if your project needs a Python server or scheduled jobs.
- **Consider a coding agent** once the project is big and every message is expensive: our [Cursor vs Claude Code vs Codex](https://vibe-code.blog/en/cursor-vs-claude-code-vs-codex) comparison helps you choose.

To decide between the three best-known app builders, read [Lovable vs Bolt vs Replit](https://vibe-code.blog/en/lovable-vs-bolt-vs-replit). And to try it for free, open an account on [Bolt](https://bolt.new/) and run the prompt above.

## Frequently asked questions

## Frequently asked questions

### Is Bolt.new free?

Yes. Bolt.new has a free plan with no card required, giving 1 million tokens a month with a daily limit of 300K tokens. It includes hosting and databases, with a Bolt badge on published sites. Free tokens do not roll over, and a free site that goes past 10 GB of bandwidth or 333,333 requests in a month stops serving content until the next cycle.

### How far do 10 million tokens go?

There is no fixed answer, because usage depends mostly on project size: Bolt rereads and syncs your files with every message. A small prototype uses little per message, a large project much more. To stretch your tokens, make small edits by hand in Code view, which costs nothing, and keep requests focused on one feature.

### What is the difference between Bolt and Lovable?

Both are browser-based app builders with hosting and a database. Bolt charges in tokens, whose cost rises with project size, and keeps the code and terminal in view. Lovable charges in credits based on request complexity and leans more towards non-technical users. Bolt also handles mobile apps through Expo.

### Can you export the code from a Bolt project?

Yes. Bolt integrates with GitHub for version control, backups and collaboration, and you can also import an existing repository. The docs state that you own your code and can use it for any legal purpose. Connecting GitHub early means you can carry on with another tool whenever you need to.

### Which Bolt agent should you use, Standard or Max?

Standard is built for everyday work: small and medium apps, interface updates and well-defined tasks. Max targets large projects, interconnected features, refactors and open-ended tasks. Only Standard is available on the free plan. Keep Max for genuinely complex requests to save tokens.

## Sources

- [Plans & pricing — Bolt](https://bolt.new/pricing)
- [Intro to Bolt — Bolt Help Center](https://support.bolt.new/get-started/intro-bolt)
- [Tokens — Bolt Help Center](https://support.bolt.new/account-and-subscription/tokens)
- [Billing — Bolt Help Center](https://support.bolt.new/account-and-subscription/billing)
- [Choose an agent — Bolt Help Center](https://support.bolt.new/building/using-bolt/agents)
- [Add and manage a built-in database in your Bolt project — Bolt Help Center](https://support.bolt.new/docs/database)
- [Bolt Cloud hosting plans — Bolt Help Center](https://support.bolt.new/bolt-cloud/hosting/plans)
- [Expo for mobile apps — Bolt Help Center](https://support.bolt.new/integrations/expo)
- [Release notes — Bolt Help Center](https://support.bolt.new/release-notes)

— Thierry Gustin, Formateur en intelligence artificielle (GustinAI) · November 21, 2026
