How to Use Bolt: The Beginner's Guide (First Prompt, Plan Mode, and a Live App)
CodexFuse Team · 2026-10-28T11:37:19.065Z · 15 min read
Learn how to use Bolt.new to build your first app without writing code: Plan mode, strong prompts, the free token budget, Supabase, and publishing — in one beginner guide.
Learning how to use Bolt takes about ten minutes, and you do not need to write a single line of code to get there. Bolt.new is StackBlitz's AI app builder: you describe the app you want in plain English, it writes the code, runs the app in your browser tab, and deploys it to a live web address. This guide walks through the whole path, from creating a free account to publishing your first app, and it covers the two things beginners get wrong most often: how the free token budget works, and why you should build one feature at a time instead of everything at once.
If the idea of describing software instead of writing it sounds familiar, that is vibe coding: you stay in charge of what the app should do, and the AI does the typing. Bolt is one of the most popular tools in that lane. Our Lovable beginner guide covers the other big builder in the same lane, and this guide covers Bolt, so you can pick the one that fits how you like to work. This is the bolt.new for beginners path: account, Plan mode, first prompt, layers, and a live deployment, in that order.
What Is Bolt.new?
Bolt.new is a free, browser-based AI app builder from StackBlitz that turns a plain-English description into a working full-stack web app. You type a prompt, Bolt scaffolds the project, installs the packages, starts a development server, and shows a live preview, all inside one browser tab. The trick is WebContainers: the Node runtime, package manager, file system, and web server are compiled to WebAssembly and run inside your browser, so there is no setup, no install, and no cold start. Bolt is one of the three most-used AI app builders, alongside Lovable and v0, according to a 2026 comparison by ToolFreebie.
Bolt builds real code, not a locked-in format. The project it scaffolds is a normal web app, usually React, Next.js, Vue, Svelte, or Astro with Tailwind, and you can download it as a ZIP file or push it to GitHub and keep working on it anywhere. That matters the moment your app outgrows the free plan, because you are never trapped inside Bolt.
How Do You Use Bolt.new?
To use Bolt.new, create a free account, open a new project, and describe the app you want in plain English. Bolt builds the first version, you test it in the preview and ask for changes one at a time, then you deploy it to a live URL.
- Create a free account at bolt.new with GitHub, Google, or email. No credit card is needed for the free plan.
- Start in Plan mode and talk the app through before anything is built. Plan mode plans without changing your project, and it uses fewer tokens than building.
- Write your first prompt as one complete description of the app: what it is, who uses it, the pages it needs, and what data it stores. A specific prompt produces a working scaffold in a single pass.
- Check the preview as soon as the build finishes. Click through every page on a laptop and on a phone before asking for the next change.
- Add features one at a time, one prompt per feature, testing and saving a version after each one. Never bundle a booking form, payments, and logins into a single prompt.
- Connect a database when the app has to remember anything between visits. Bolt can connect Supabase and set up a free project for you from inside the chat.
- Deploy with the built-in Netlify deploy, which takes about twenty seconds, and share the live link. A custom domain and removing Bolt branding need the Pro plan.
What Should Your First Prompt Say?
Your first prompt should name the app, its users, its pages, its data, and its style in one complete message. Vague prompts force follow-ups, and every follow-up resends the whole conversation and file history, so each one costs tokens. One specific prompt beats five clarifying ones, because the cost of a conversation grows with chattiness.
A strong first prompt has four ingredients: the app in one sentence, the pages it needs, the data it stores, and the look you want. Here is one you can adapt, for a small booking app:
Example first prompt: "Build a booking app for my three-person house-cleaning company. Pages: Home, Services and prices, Book a cleaning, and Crew. Customers pick a service, a date, and a morning or afternoon window, then enter their name, phone, and address. Store bookings with a status of requested, confirmed, or done. Crew members log in and see only their own jobs. Use a clean, simple design with navy blue and white."
Notice what the prompt does not do: it does not ask for payments, email reminders, or reports. Those are later features, each with their own prompt. The first prompt builds the scaffold and the core idea only. Bolt's own blog puts it plainly: build the feature your app exists for first, then add the rest one at a time.
What Is Plan Mode, and Why Should You Use It First?
Plan mode is Bolt's way of talking through your app before any code is written. You describe what you want, and Bolt asks questions, lays out the pages, the data, and the build order, and shares a plan. Once a project exists, Plan mode makes no changes to it, and it uses fewer tokens than building, according to Bolt's own documentation. Starting from the homepage, Plan mode sets up your app's base structure first, then shares the plan.
Use Plan mode for the conversation that decides what to build. Settle three questions there: what the app is for, who uses it, and which feature comes first. Then switch to Build mode and let Bolt execute the plan one layer at a time. A plan is not a contract, either. If a new idea changes your data or who logs in, update those layers first and test the features that sit on them before adding anything new. If the change only adds a feature, add it as the next layer like any other.
Beginners skip Plan mode because it feels like a delay. It is the opposite: ten minutes of planning saves the tokens you would otherwise spend rebuilding a booking form that was written before anyone decided what a booking contains. That is the whole point of bolt.new plan mode: think first, build second, and spend tokens only on a plan you already believe in.
How Do You Build Your App One Feature at a Time?
You build your app one feature at a time by working in layers, from the parts every feature depends on to the parts nothing depends on. Bolt's own blog, published October 2, 2026, gives the order: plan first, then pages, data, and logins, then features one at a time, then polish, then publish. Each layer settles a question the next layer needs answered.
| Layer | What you build | Why it comes here |
|---|---|---|
| 1. Plan | The app's purpose, users, pages, data, and first feature, in Plan mode | Every later decision depends on it |
| 2. Pages | Core pages and navigation, placeholder text is fine | Every feature needs a place to live |
| 3. Data | The tables: customers, services, bookings, and example rows | Every feature reads or writes this data |
| 4. Logins | Who signs in and what each role can see | Adding privacy rules later means re-checking every feature |
| 5. Core feature | The one thing the app exists to do, such as the booking form | This is the app; everything else supports it |
| 6. Next features | One feature per prompt: reminders, confirmations, reports | One change means one suspect when something breaks |
| 7. Polish and publish | Colors, wording, spacing, then a live deployment | Restyling a finished page is cheap; restyling a moving target is not |
The reason for the discipline is simple. Add a booking form, a payment step, and staff logins in one prompt, and when the form breaks you cannot tell whether the form itself is wrong or one of the other changes broke it, because three things changed at once. Bolt's documentation states the rule directly: "Do one thing at a time. Don't try to add multiple features in one go." Google Cloud's DORA research program found the same habit in professional teams, calling small batches easier to triage and remediate, and a safety net for AI adoption.
Each layer is done when it passes three checks, and then you save the version before starting the next one: use it the way a customer would, start to finish; do it again on a phone; and try the wrong input on purpose, such as a blank phone number or a date in the past. The wrong-input check finds the problems a normal run-through will not. If the booking form accepts a cleaning for last Tuesday, you want to find that before three more features depend on the form.
What do you do when a new feature breaks something?
When a new feature breaks something, roll back to the last version that worked, then try again with a better description. Bolt's Version history restores your project to a previous state without consuming tokens, so a bad change costs you nothing to undo. One limit: restoring a version does not roll back your database, so if the broken feature changed your data, check the data too. Do not keep clicking Attempt fix, either. Each attempt uses tokens, and restoring is free. Describe what went wrong in your own words, such as "the confirmation email goes out before the booking saves," and rebuild that one feature. Because only one thing changed, you know which thing to fix.
How Far Does the Free Plan Actually Go?
The free plan goes far enough to build and publish a real first app, but it has a hard daily ceiling you need to plan around. Bolt's pricing page, checked October 28, 2026, gives the free plan 1,000,000 tokens a month with a 300,000-token daily cap, no credit card required. The monthly allowance resets on the 1st of each month and the daily cap resets every day, and free tokens never roll over. Pro costs $25 a month for 10,000,000 tokens, with rollover, your own domain, no Bolt branding, and access to the newest AI models. Teams costs $30 a month per member.
| Plan | Price | Tokens per month | Daily cap | What else |
|---|---|---|---|---|
| Free | $0 | 1,000,000 | 300,000 | Public and private projects, Bolt branding, website hosting |
| Pro | $25 a month | 10,000,000 | None | Own domain, no branding, rollover, newest models, SEO boosting |
| Teams | $30 a month per member | 10,000,000 per member | None | Everything in Pro, plus shared capacity and team controls |
Tokens are the real budget, and the daily cap is the real constraint. Each prompt resends the entire file tree, conversation history, and tool definitions to the AI, so a medium edit to a five-file project, such as adding a search bar that filters a post list, typically burns 30,000 to 60,000 tokens, according to ToolFreebie's guide, which flags these as estimates from Bolt's docs and third-party write-ups rather than measurements. An independent 2026 comparison pegs realistic output at roughly 3 to 8 meaningful prompts a day. That is enough to scaffold a small project and refine it, not enough to ship a real app in one sitting.
Here is what one 300,000-token day buys, using those published estimates. A detailed first prompt that scaffolds a small app costs roughly 50,000 to 80,000 tokens. Three medium edits at about 45,000 tokens each cost about 135,000. A database connection and a login feature together cost roughly another 80,000. The total lands near 265,000 tokens, inside the daily cap, with a little room for one fix. In short, bolt.new pricing for beginners comes down to one question: does the bolt.new free plan cover your build, or does the daily cap keep stopping you? The lesson is the same as the build order: front-load the detail into fewer, bigger prompts, fix small things in the editor by hand instead of spending 20,000 tokens on a typo, and export the project to GitHub when the day's budget runs out.
How Do You Deploy a Bolt.new App?
You deploy a Bolt.new app with the built-in Netlify deploy, which takes about twenty seconds and gives the app a real public URL. That is how to deploy a bolt.new app in under a minute, and the same button works again after every new layer you add. Bolt runs the app in your browser while you build it, and deploying moves it to hosting other people can reach. Custom domains and removing Bolt branding are paid features, so a free deployment carries Bolt branding and a Bolt web address.
Deploy when the features you need on day one pass their checks and you have done a polish pass. You do not need every feature first: on Bolt.new you can keep adding layers to a live app, one feature at a time, testing and saving a version after each one. When the app matters enough to leave Bolt's hosting, download the full project as a ZIP file or push it to GitHub from inside Bolt, then host it anywhere. If you would rather keep the app running on your own server, our best VPS for n8n guide shows how VPS hosting works, and the same idea applies: a small server you control, with the code you own.
If the token ceiling is your problem rather than the hosting, the open-source bolt.diy project is the escape route. It ships the same chat-and-preview interface but lets you bring your own API key, so you can pair it with a provider that offers a free tier, such as Gemini or Groq. Setup takes about thirty minutes, and the provider's own rate and usage limits still apply, so check those before starting a large project.
Bolt.new vs Lovable: Which Should a Beginner Pick?
Bolt.new and Lovable are the two most popular AI app builders for beginners, and they cost the same at the paid level, $25 a month each, but they feel different to use. Lovable targets non-developers who want polish over control: it produces the most polished output with the least input. Bolt is far more developer-friendly. Its editor is a real code editor, the project is a normal codebase you can download and read, and iteration is instant because the runtime runs in your browser tab instead of a remote sandbox.
| Bolt.new | Lovable | |
|---|---|---|
| Runs in | Your browser tab (WebContainers) | Browser, remote sandbox |
| Free plan | 1M tokens a month, 300K a day | 5 build credits a day, up to 30 a month |
| Paid plan | Pro, $25 a month, 10M tokens | Pro, $25 a month |
| One-click deploy | Netlify, built in | Built in |
| Database | Connect Supabase | Connect Supabase |
| Best for | Builders who want the code and fast iteration | Non-developers who want polish with least input |
| Self-hosted option | Yes, bolt.diy | No |
The comparison figures come from ToolFreebie's 2026 comparison table, which re-checked the free-plan details against the vendors' pricing pages on October 24, 2026. Pick Lovable if you want the most polished first version with the least prompting, and read our Lovable beginner guide to start there. Pick Bolt if you want to see and edit the code, iterate instantly, and keep a self-hosted escape route. If you want to compare a third lane, the prompt-first builder against a model playground, our Google AI Studio beginner guide covers the other major free AI-builder lane.
What Bolt.new Cannot Do (Yet)
Bolt.new cannot do everything, and knowing the edges before you start saves a wasted afternoon. WebContainers run a Node-shaped runtime in your browser, so native binary dependencies do not work: no Docker inside Docker, no native Postgres binary, and no Python data-science stack with NumPy or Pandas. Some Python works through Pyodide, but it is slow and limited. For React, Next.js, Vue, Svelte, Astro, Vite, Tailwind, and TypeScript prototyping, none of that matters. For machine-learning pipelines or heavy native dependencies, Bolt is the wrong tool.
The second limit is time. A WebContainer lives only while the tab is open, so there is no daemon, no cron job, and no queue worker running in the background. An app that needs a worker checking something every hour, or a long-running background process, fits a real server better, and a tool with a real Linux virtual machine, such as Replit Agent, fits that job better than Bolt. Bolt is best for landing pages, internal admin tools, prototypes, indie software-as-a-service minimum viable products, and hackathon submissions.
Beginner Mistakes That Waste Your Free Tokens
Most beginners do not run out of ideas on Bolt. They run out of tokens, and the same four mistakes cause most of it. First, vague first prompts. A prompt that says "build me an app for my business" produces a generic scaffold, and every correction after it resends the whole project history. Write the tight design first, then prompt once. Second, building everything at once. A prompt that adds a booking form, payments, and logins together burns tokens on three features and leaves you unable to tell which one broke. Third, clicking Attempt fix again and again. Each attempt uses tokens, while restoring the last working version is free. Restore, describe the problem in your own words, and rebuild the one feature. Fourth, chatting instead of editing. Bolt's editor is real: fix a typo or a Tailwind class by hand instead of spending 20,000 tokens asking the AI to do it. Use Discussion mode for architecture talk, which carries a smaller context than a build prompt.
Sources
- Bolt.new Pricing (bolt.new/pricing) — Free $0 with 1M tokens a month and a 300K daily limit; Pro $25 a month with 10M tokens and rollover; Teams $30 a month per member; Enterprise custom. Checked October 28, 2026.
- Bolt.new Blog, "Build Your App One Feature at a Time" (bolt.new/blog/build-app-one-feature-at-a-time) — Plan mode uses fewer tokens and makes no changes; build order plan, pages, data, logins, core feature, polish, publish; "Do one thing at a time" documentation quote; Version history restores without consuming tokens; Attempt fix uses tokens. Published October 2, 2026.
- ToolFreebie, "Bolt.new: Free AI App Builder in Your Browser" (toolfreebie.com/bolt-new-free-app-builder) — WebContainers explained; free tier 1M tokens a month with a 300K daily cap, 3 to 8 meaningful prompts a day; Pro $25 a month with 10M tokens; Netlify deploy in about 20 seconds; Supabase connection; bolt.diy self-hosted option; Bolt vs Lovable vs v0 comparison table. Free-plan details re-checked against Bolt's pricing page on October 11, 2026; competitor prices re-checked October 24, 2026.
- Google Cloud DORA Research Program (dora.dev) — working in small batches makes problems easier to triage and remediate, described as a safety net for AI adoption. Cited by the Bolt.new blog, 2025.
Frequently Asked Questions
Is Bolt.new really free?
Yes. Bolt.new's free plan costs $0, needs no credit card, and includes 1,000,000 tokens a month with a 300,000-token daily cap, public and private projects, unlimited databases, and website hosting. The limits are real: deployed sites carry Bolt branding, there is no custom domain, and you cannot lift the daily cap without upgrading. It is genuinely usable for prototyping and learning; sustained daily work will hit the cap.
Do I need to know how to code to use Bolt.new?
No. The chat-box prompt is the whole interface for a first project like a booking app or a habit tracker. Reading the generated code is optional, and it is mainly useful once you want to understand or tweak something Bolt cannot fix through chat alone. Bolt is more developer-friendly than Lovable, so reading code helps more here, but it is never required to build and publish a first app.
Which AI model does Bolt.new use?
Bolt no longer names one. As of October 2026 it offers two agents, Standard and Max, and handles model selection behind the scenes. The free plan gets Standard only. You pick the depth and cost trade-off, not the provider. Pro unlocks access to the newest AI models.
Can I export my Bolt.new app and host it somewhere else?
Yes. Because the project is real code rather than a Bolt-only format, you can download the full project as a ZIP file or push it to GitHub, then deploy it through a different host such as Netlify or a VPS. That is also the practical way to keep working once the free token budget runs out: scaffold in Bolt, export, and continue locally with a free API key.
What happens when I hit the daily token limit?
Bolt stops taking build prompts until the 300,000-token daily cap resets the next day. The monthly 1,000,000-token allowance resets on the 1st of each month, and free tokens never roll over. When you hit the cap, export the project to GitHub and continue in a local tool, plan your next features offline so the next day's prompts are tight, or upgrade to Pro, which has no daily cap and 10,000,000 tokens a month.