Is Bolt.new Free? What Tokens Really Cost
The free tier is real but capped: 300K tokens a day and 1M a month — roughly 5–10 meaningful prompts once your project grows. Pro is $25/month starting at 10M tokens with rollover, Teams is $30 per member with tokens that are not shared, and the cheapest long-run strategy is to export the code and keep iterating locally with your own key.
AICoding score
Quick Verdict
8.2 / 10 — the fastest browser-based AI full-stack builder: WebContainers run the whole app locally in a tab, and the free tier is the most usable in the category. The catch is that token spend scales with project size, not with how much you prompt.
Bolt is the cheapest way to get from a prompt to a running full-stack app without installing anything — and the most transparent about what it charges for, because tokens are a bill you can watch. Where it surprises people is project growth: every request resends the whole file tree, so the same edit costs multiples more as the app gets bigger.
Use Bolt.new if…
- You want a running full-stack app from a prompt in minutes, with nothing installed
- You will export the code rather than rebuild inside the tool forever
- You want a free tier you can genuinely evaluate on before paying
- You want hosting elsewhere — export to Netlify/Vercel and pay $0 for hosting
Skip Bolt.new if…
- One large project is your whole workflow — token spend grows with the file tree
- You need native binaries (Docker, native Postgres, ML pipelines): WebContainers cannot run them
- You are paying for a team where only a couple of people actually build (tokens are per member)
Core Features
WebContainers (in-browser runtime)
The Linux runtime, npm, file system, and HTTP server run inside your browser via WebAssembly. Zero cold start — the dev server is ready the moment the agent finishes writing files, and compute is your laptop, not a cloud VM. This is why a $25 Pro plan can bundle 10M tokens.
Full-Stack Generation
One prompt generates frontend, backend APIs, database schemas, and server logic. Unlike v0 (frontend only) or Lovable (React + Supabase only), Bolt builds a complete runnable application end-to-end.
Fast Edit-Run-Fix Loop
The browser-local runtime makes iteration feel instant: edit, preview, fix — no cloud sandbox reload between changes. The closest thing to "pair-programming with an AI" in a browser tab.
Complete Code Ownership
Download a full-project zip or sync to GitHub. Export and continue in Cline/Aider/local editors with your own key — no lock-in, no token pressure. This is the biggest difference from "demo toy" builders.
bolt.diy (open-source clone)
A community-maintained open-source version of Bolt you can self-host with your own API key — escape the token ceiling entirely by running Bolt's engine locally against free Gemini/OpenRouter models.
Pricing
Public and private projects, hosting, unlimited databases, up to 333K page requests a month — but deployed sites carry Bolt branding, uploads cap at 10MB, and the daily token ceiling lands fast on a growing project.
- 300K tokens per day
- 1M tokens per month
- Hosting + unlimited DB
- Bolt branding on sites
No daily token cap and a 10M-token starting allowance, branding removed, private sharing, custom domains, SEO controls, 100MB uploads and up to 1M page requests. Unused tokens roll over for one extra month while the subscription stays active.
- 10M tokens to start
- No daily cap
- Rollover (1 extra month)
- Custom domain + SEO
Everything in Pro plus centralised billing, team access management, provisioning controls, organisation-wide sharing, private NPM registry support and design-system knowledge. Important: each member gets their own token allowance and tokens are not shared.
- Per-member tokens (not pooled)
- Central billing
- Access management
- Private NPM registry
Everything in Pro plus SSO, audit logs, compliance support, data governance and retention policies, a dedicated account manager, 24/7 priority support, custom workflows, SLAs and flexible procurement. No price is published.
- SSO + audit logs
- Data governance
- Dedicated AM, 24/7
- Custom SLA
The community open-source clone you run yourself against your own API key — including free Gemini or OpenRouter keys. No token ceiling, no vendor bill, full control; you pay in setup and a rougher editor.
- Open source
- Your own API key
- No token ceiling
- Self-hosted
* Verified against the official Bolt pricing page on September 13, 2026. Annual billing is listed at up to 28% off, but the exact annual prices are not published; top-up token pricing is also absent from the pricing page, so treat any overage figure as unknown until you see it in Stripe.
What Bolt.new Really Costs
| Task / Plan | Usage | Typical cost | What to watch |
|---|---|---|---|
| Free tier, light use | Scaffold a project and a few edits a day | $0 | 300K tokens/day and 1M/month |
| Free tier, real project | Multi-file edits on an existing app | Caps hit mid-session | Every prompt resends the whole file tree |
| Pro, regular prototyping | 10M tokens/month, no daily cap | $25/mo | Unused tokens carry one extra month |
| Pro, heavy iteration | Large project, constant rebuilds | $25 + top-ups | Top-up pricing is not published — ask before you depend on it |
| Teams, 5 developers | Per-member allowances; tokens are not pooled | $150/mo | You pay per person even if only two people generate code |
| Enterprise | SSO, audit logs, compliance, SLAs | Custom quote | No published number |
| Scaffold then export | Build in Bolt, export the zip, iterate locally on your own key | $0 tokens after export | The single biggest lever on a Bolt bill |
| bolt.diy self-hosted | Open-source clone against free Gemini/OpenRouter keys | $0 software, your keys | No ceiling — you own the runtime instead |
| Annual billing | Pay yearly instead of monthly | Listed at up to 28% off | Exact annual prices are not on the pricing page |
Bolt adds no markup on top of its own token pricing, but token spend tracks project size rather than effort — the same edit costs more in a bigger file tree. Verified against the official pricing page, September 13, 2026; top-up and annual prices are not published, and are marked as unknown rather than estimated.
Bolt.new Deals & Ways to Pay Almost Nothing
No coupon codes here — the real levers are the free tier, exporting early, and running the open-source clone on your own key.
vs The Field
Bolt.new vs Lovable
Developer control vs non-technical polishBolt runs the full stack in-browser (WebContainers), supports more frameworks, and gives you readable, editable code. Lovable targets non-technical founders: built-in Supabase, one-click deployment, and polish over control. If your team reads code: Bolt. If the team is non-technical and output looks good matters most: Lovable. This is the most-searched comparison in the category.
Bolt for devs · Lovable for foundersBolt.new vs v0 (Vercel)
Full-stack vs frontend componentsv0 generates polished React/Next.js UI components (shadcn/ui, Tailwind) — frontend only. Bolt builds complete full-stack apps with backend, database, and server logic. Landing page or UI components: v0. Working full-stack app: Bolt. They are complementary — many teams use v0 for UI and Bolt for the API layer.
Complementary, not rivalsBolt.new vs Replit Agent
In-browser speed vs cloud autonomyReplit runs a cloud VM with 50+ languages, built-in hosting/database, and the most autonomous agent. Bolt runs in-browser (Node ecosystem) with near-instant iteration and more predictable token pricing, but needs export to deploy. Language flexibility + built-in hosting: Replit. Speed + cost predictability: Bolt.
Bolt for speed · Replit for autonomyWhat It Is Worth Paying For
Prompt-to-prototype
From "build a Next.js todo app with Supabase auth" to a running app in under a minute — the fastest way to validate an idea in the browser.
Frontend + backend scaffold
One prompt generates a full-stack app with APIs and database setup, unlike frontend-only tools. The scaffold-to-production foundation.
Mobile prototypes via Expo
The only major AI builder producing React Native/Expo mobile apps — extend your web prototype to mobile in the same tool.
Scaffold then export
Use Bolt for structure, then export the zip to Cline/Aider with a free key and iterate without token pressure. The recommended power workflow.
Self-host with bolt.diy
Escape token ceilings entirely by self-hosting the open-source clone with free Gemini/OpenRouter keys — full control, unlimited iteration.
The Cheapest Way to Use Bolt
Size your project on the free tier
The 300K/day cap is the fastest way to learn what your idea costs in tokens: build the real thing on the free tier for a day and watch the meter, rather than guessing from demos.
Export once the scaffold is stable
Bolt is at its best generating structure. The moment the app runs and the shape is right, download the zip or sync to GitHub — every later edit inside Bolt costs more because the file tree is bigger.
Continue locally on your own key
Iterate in Cline, Aider or bolt.diy with your own provider key: pay-as-you-go instead of a token ceiling, no branding, and the code was always yours. This is the path that turns a $25/month habit into a few dollars of API usage.
FAQ
Is Bolt.new free?
Yes — Bolt has one of the most generous free tiers among AI app builders: 1M tokens per month with a 300K-per-day cap, which realistically delivers 5-10 meaningful prompts a day. That is enough to scaffold a small project and refine it. The free tier has two limits: Bolt branding on deployed sites and public projects only (no private). To escape the token ceiling, scaffold in Bolt then export the code and continue in a free CLI agent with your own key.
How much does Bolt.new cost?
Free tier: $0 with 1M tokens/month. Pro: $25/mo starting at 10M tokens with rollover (unused tokens carry one extra month while the subscription stays active). Teams: $30 per member per month — note that tokens are per member and are not pooled. Enterprise: custom with SSO. The token-based model is more predictable than Replit's effort-based billing, but complex prompts burn tokens fast — a medium edit to a 5-file Next.js app can consume 30-60K tokens.
Is Bolt.new better than Lovable?
They serve different users. Bolt is developer-friendly: it runs the full stack in your browser via WebContainers, supports more frameworks (React, Vue, Svelte, Angular, Astro, Expo), and gives you full code ownership you can read and edit. Lovable is built for non-technical founders: built-in Supabase, one-click deployment, and polish over control. If you want to read/edit the code yourself: Bolt. If the team is non-technical and output polish matters most: Lovable.
Is Bolt.new better than v0?
They are complementary. v0 (by Vercel) generates polished React/Next.js frontend components with shadcn/ui and Tailwind — frontend only. Bolt generates complete full-stack applications including backend APIs, database setup, and server logic. If you need a landing page or UI components: v0. If you need a working full-stack app with a Node.js backend: Bolt. Many developers use v0 for UI and Bolt for the API layer.
How does Bolt.new compare to Replit Agent?
Both are browser-based full-stack AI builders, but differently. Replit runs on a cloud VM supporting 50+ languages with built-in hosting and database; Bolt runs in-browser via WebContainers (Node.js ecosystem, WASM) with near-instant iteration but requires export to deploy. Replit's Agent is more autonomous on long-running tasks; Bolt's edit-run-fix loop is faster and its token pricing more predictable. Language flexibility + built-in hosting: Replit. Speed + predictable cost: Bolt.
What is a WebContainer?
WebContainer is StackBlitz's technology that runs a Node.js-compatible Linux environment entirely inside your browser using WebAssembly. The Node runtime, npm, file system, and HTTP server all run in the tab — no remote server. This means zero cold start (the dev server is ready the moment code is written), and your compute is local. The trade-off: WebContainers cannot run native binaries not compiled for WASM — no Docker-in-Docker, native Postgres, or NumPy/Pandas for ML pipelines.
What frameworks does Bolt.new support?
React, Next.js, Astro, Vue, Svelte, SvelteKit, Remix, Vite, and plain HTML/CSS/JS, plus Node.js backends with Express/Fastify. It is also the only major AI builder that can produce React Native/Expo mobile apps. This multi-framework flexibility is Bolt's main differentiator over v0 (React/Next.js only) and its broadest advantage in the category.
Can I export my Bolt.new project?
Yes — the "Download" button ships a full-project zip with complete code ownership, and GitHub sync is built in. A common workflow: use Bolt to scaffold + structure, export the code, then continue iterating in Cline, Aider, or your local editor with a free Gemini/OpenRouter key — no token budget pressure. For production, export and deploy to Netlify, Vercel, or any host.
Why does Bolt.new run out of tokens so fast?
Every prompt resends the entire file tree, conversation history, and tool definitions to Claude. A medium-complexity edit to a 5-file Next.js project ("add a search bar that filters the post list") typically consumes 30-60K tokens of context. On the free tier's 300K daily cap that is 5-10 meaningful prompts. Two workarounds: plan offline first (tight design doc before opening Bolt), and use Bolt for scaffold then export to a local agent.
Is Bolt.new good for production apps?
For getting to a working prototype fast: yes, excellent. For production hardening: Bolt generates production-quality code you can export to GitHub and deploy anywhere, but complex apps with custom auth, intricate database schemas, or heavy backend logic typically need developer review and refinement. It excels at the scaffold-to-prototype phase; production usually means manual work on top.
Why do I run out of tokens even on Pro?
Because token spend is driven by project size, not by how many prompts you send. Every request resends the file tree, conversation history and tool definitions, so a medium edit in a 5-file Next.js app runs 30-60K tokens, and the same edit in a 40-file app costs multiples of that. On Pro (10M tokens starting allowance) that is comfortable for a handful of projects and tight for one large app you rebuild all day. The two levers that actually work: keep projects small and start new ones instead of growing one monolith, and export the code once the scaffold is stable.
Is the Teams plan worth it if tokens are not shared?
Only if every member you pay for is building. Teams is $30 per member per month and the token allocation is per member — there is no shared pool, so paying for five people when two are generating code buys three unused token allowances. What Teams adds on top of Pro is administrative: centralised billing, team-level access management, provisioning controls, private NPM registry support and design-system knowledge. If your need is collaboration rather than governance, keep one Pro seat and export the project instead.
How do I avoid the token bill altogether?
Treat Bolt as a scaffold factory. Build the structure on the free tier or one Pro month, then download the project zip (or sync to GitHub) and continue in a local CLI agent — Cline, Aider or bolt.diy — with your own API key. Token spend switches to pay-as-you-go provider rates you control, there is no ceiling, and the code is yours. bolt.diy, the open-source clone, goes further: self-host it against free Gemini or OpenRouter keys and the software cost is zero. The honest trade is setup work and a less polished editor.