How do you take a Bolt.new app to production?
You take a Bolt.new app to production by deciding where it will be hosted, moving every secret out of browser code, and making sure your data lives in a secured database you control. Most other launch work, such as analytics and legal pages, can follow once those three are solid.
Bolt makes the first version fast, and its built-in Bolt Cloud now covers hosting, databases and domains inside the editor. The gap is everything a prototype doesn't need: environments, access rules and a way to recover when something goes wrong. This guide is the Bolt-specific version of those steps. For where it fits in the wider build, see our guide to MVP design and development.
What breaks when a new Bolt app goes to production?
When a new Bolt app goes to production, the first things to break are usually the hosting setup, exposed secrets and an unsecured database. They work fine for one person in a preview and fail once strangers, search engines and real data arrive.
Area | Prototype default | What goes wrong | Production fix |
|---|---|---|---|
Hosting | Public bolt.host link | Wrong domain, indexed test pages, no rollback plan | Custom domain, GitHub sync, chosen host |
Secrets | Keys pasted into chat or code | Keys shipped in the browser bundle | Server functions reading stored secrets |
Database | Created automatically | Open tables, no backups, test and real data mixed | Security audit, RLS, owned database, separate environments |
Changes | Edits go live on Update | A prompt breaks the live app | Branches and a staging copy |
If you haven't decided which features belong in version one, do that first, because every extra table and integration adds to this list. Our guide on how to scope an MVP covers cutting features without cutting the point.
Should you host a Bolt.new app on Bolt, Netlify or your own host?
Host on Bolt if your app uses Bolt's database and server functions and you want the simplest path. Move to your own host through GitHub when you need preview environments, team code review or a stack your developers already run.
- Bolt hosting. Bolt's docs say Publish puts your app on a free bolt.host address, and paid plans can attach a custom domain. Sites are public and indexable by default; private sites need a paid plan. Changes go live only when you click Update.
- Netlify from Bolt. Bolt's docs list Netlify as a publishing option, but only for apps that run entirely in the browser. Apps that need Bolt's database or backend should stay on Bolt hosting.
- Your own host through GitHub. Bolt syncs with GitHub, and from the repository you can deploy to any host that builds your framework, such as Vercel. Our Vercel integration page covers how we set that up.
Bolt's GitHub docs describe a two-way sync: Bolt commits your changes automatically and checks GitHub every 30 seconds for outside edits, and if both change at once, Bolt's version wins. Merging branches has to happen on GitHub. Those details matter once a developer starts working in the repository alongside the editor.
How do you handle environment variables and secrets in Bolt.new?
Handle secrets in Bolt.new by storing them in the project's Secrets settings and reading them only from server functions, never from code that runs in the browser. Bolt's docs say secrets let server functions call outside services without exposing keys to users.
The common leak is frontend config. Many Bolt projects are Vite and React apps, and Vite's documentation says variables prefixed with VITE_ are exposed to client code, which means they end up in files anyone can read. Next.js treats the NEXT_PUBLIC_ prefix the same way. A payment key or an AI provider key behind either prefix is public, whatever the variable is called.
- Split public from private. Public values, such as a Supabase project URL and its public anon key, can sit in client code because row-level security protects the data. Service-role keys, payment secret keys and AI keys can't.
- Rotate anything that leaked. If a key was ever pasted into a prompt, committed to a repo or shipped in client code, replace it at the provider, then store the new one as a secret.
- Set values per environment. If you deploy elsewhere, add variables in your host's settings for production and preview separately, so a test environment never touches live payments.
How do you set up a real database for a Bolt.new app?
Set up a real Bolt.new database by securing the one Bolt created, then deciding whether you need to own it directly in Supabase. Either way, add access rules, backups and a separate development copy before real users sign up.
Bolt's docs say Bolt creates a database automatically when your project needs one, and offer a security audit that looks for open permissions and missing row-level security policies. Run it, and fix every finding before launch. Note one detail from the same docs: unpublished projects may have their database paused after six or more days of low usage, so don't treat an unpublished preview as a live service.
If you want full control, Bolt's Supabase docs describe a Claim option, on Pro or Teams plans, that moves a Bolt database into your own Supabase account. They warn that connecting a separate Supabase project to an app that already has a Bolt database replaces the connection and may lose data, so claim instead. Once it's in your account, check Supabase's backup options and create a separate project for development.
For example, suppose a founder launching a Bolt-built client portal
Hypothetical example: a founder has built a small client portal in Bolt with sign-in, a projects table and file uploads. The preview works, and they want ten paying customers on it next month.
They run the database security audit and find a table any signed-in user can read; they add a policy limiting rows to their owner. They find an AI provider key in a frontend file, rotate it at the provider and move the new one into Secrets behind a server function. They claim the database into their own Supabase account, connect GitHub, attach a custom domain and create a branch for future changes so prompts stop landing straight on the live app.
Nothing here is a measured result; it shows the order of work.
What's the step-by-step checklist for a Bolt.new app in production?
Work through the steps below in order, because each one protects the next. Security and data come before the domain, since a polished address on an open database just sends more people to the leak.
Connect GitHub
Link the project to a private repository so you have history outside Bolt, and start making changes on a branch instead of main.
Audit secrets
Search the code for API keys, rotate any that were exposed, and move them into Secrets read by server functions.
Secure the database
Run Bolt's security audit, add row-level security on every table and test it with two accounts that shouldn't see each other's data.
Decide who owns the data
Keep Bolt Database or claim it into your own Supabase account, then set up backups and a separate development database.
Choose hosting and domain
Publish on Bolt with a custom domain, or deploy the repository to a host such as Vercel with production and preview environments.
Add monitoring and analytics
Add error tracking and basic product analytics, so you hear about failures before your users report them.
Test the core flow on a phone
Sign up, complete the main task and pay, if there's billing, on a real device with a fresh account.
When should you move a Bolt.new app off Bolt entirely?
Move a Bolt.new app off Bolt when you need things the editor isn't built for: code review, automated tests, several environments and more than one developer working at once. That point usually arrives with paying customers, not before.
The move is simpler than it sounds because the code is already in GitHub. The harder part is cleaning up what fast prompting leaves behind, which our guide to rescuing a vibe-coded app covers in order of risk. If you're weighing whether to keep prompting or bring in people, our comparison of vibe coding vs hiring a team sets out the trade-offs, and our MVP design service is the version where we design and build it with you.
One more case: if your marketing pages live inside the same client-rendered app, search engines and AI answers may struggle to read them. Our guide to splitting your marketing site from your app explains the options.
