Bolt.new
Vibe-code an entire web app — frontend, backend, database — from one prompt.
You have an idea for an app you'd use once a week — a workout tracker, a reading log, a class attendance pad, a tip-splitter for your Friday dinner group. It's not a startup. You won't hire a developer. You just want it to exist by Saturday.
Why this tool matters
Bolt.new is StackBlitz's flagship generative-app tool: you describe an app in English, and it writes the frontend, backend, database, and deploys it to a live URL — all inside a browser tab. No local environment, no terminal, no deployment ceremony. The gap between “I wish this app existed” and “it exists at a URL I can share” has collapsed to about twenty minutes.
This is the tool that makes real the promise of “vibe coding” — building software by describing what you want rather than writing it. Bolt sits at the full-stack end of the spectrum: while v0 (Day 9) produces individual components, Bolt produces the whole working application, with routes, API endpoints, a database schema, authentication, and a deploy to StackBlitz's infrastructure or Netlify.
For everyone who has ever said “I'd build that if I could code” — and for many who can code, but want to skip the setup — Bolt is a genuine unlock. It is not a replacement for craft software engineering, but for small tools, internal dashboards, one-off apps, prototypes, class projects, and weekend experiments, it compresses a three-day build into an afternoon.
Setup
Account: sign in at bolt.new with a StackBlitz account (Google/GitHub works). Free tier gives you a daily token budget — enough for a meaningful app; Pro is $20/month for heavier use.
Expectations: Bolt is stronger on frontend-heavy apps (React, Next.js) than on backend-heavy ones. A CRUD app with a handful of tables and some auth is its sweet spot. A multi-tenant SaaS with billing integration is stretching it.
Walkthrough
Step 1: Describe the app in one paragraph
At bolt.new, type a paragraph describing the app: “A personal reading log where I can add books I've read with title, author, finish date, and a rating 1-5. List view with sorting by date or rating. Simple add-new form. No auth yet; just me.”
Step 2: Watch it build in real time
Bolt streams the file creation: package.json, routes, components, database schema, styles, all appearing in the left sidebar. Within about ninety seconds you have a working app running in the preview on the right.
Step 3: Interact with the running app
Use the preview pane to try the app. Add a book. Notice that it persists across refreshes (Bolt uses SQLite or similar by default). Everything is already hooked up end-to-end. No separate server to start.
Step 4: Iterate in natural language
In the chat pane, describe changes: “Add a genre field. Filter the list by genre. Show a count of books per genre at the top.” Bolt rewrites the necessary files, restarts the app, and the preview updates live.
Step 5: Add auth when you need it
Once the app is working for you alone, add: “Add email + password authentication. Each user should only see their own books.” Bolt scaffolds auth, adds a user table, and updates every query to filter by user. Review the security assumptions — auth is where AI-generated code most often introduces bugs.
Step 6: Deploy to a live URL
Click Deploy. Bolt deploys to Netlify in one click; you get a permanent URL. Share it, bookmark it, use it. For an app you'll develop further, click Export to GitHub to move the code into your own repo and keep iterating outside Bolt.
Your turn
Basic: A personal one-user tool
Think of a tool you'd use at least once a week. Examples: a reading log, a workout tracker, a recipe box with search, a note-to-self log for a project, a simple CRM for your freelance clients. Keep it small and singular — one user, a few fields, no fancy permissions.
Build it in Bolt in under 30 minutes. Deploy it. Bookmark the URL. Use the tool this week.
Advanced: An app you'd show someone
Pick a genuine need in your community — a sign-up tool for your class, a menu-voting app for your potluck group, an inventory tracker for a volunteer organization, a Q&A board for a conference panel. Something with real users (even if just five).
In Bolt, build: the frontend, auth (even if just email-only magic links), the database schema, and at least one non-trivial feature beyond basic CRUD (search, filtering, a small chart, email notifications, CSV export).
Deploy. Invite two actual people to use it. Fix the three things they immediately complain about.
Write a 200-word postmortem: what did Bolt do well? What was the hardest part to get right (usually auth + data model choices)? What would you do differently on the next app?
This exercise is the closest thing in this course to shipping a real product. It's the one that most changes how you think about software.
Pitfalls and pro tips
Auth is where AI-generated code bites. Bolt can scaffold auth in thirty seconds, but subtle bugs (tokens not verified, passwords stored insecurely, session fixation) are common. For any app with real users, have a human review the auth code specifically before going live.
The database schema is hard to refactor. Bolt picks a schema early based on your first prompt. Changing it later (renaming a table, restructuring a relationship) often causes regressions. If you can, sketch the data model on paper before your first prompt.
Credentials leak into code. Bolt will sometimes hardcode API keys or database URLs in source files rather than environment variables. Before pushing to a public repo, search the whole codebase for anything that looks like a secret.
How it compares
Bolt.new's closest competitors are Lovable (similar vibe-coding product, slightly different house style and deployment story), Replit Agent (deeper into the IDE, with a chat-driven agent that can open and edit files), and v0 (Day 9, same company's frontend-only counterpart). For a one-shot full-stack app with deployment, Bolt and Lovable are effectively interchangeable; try both and pick the one whose output style fits your taste. For a project you intend to keep building on for months, Replit Agent's deeper editor experience tends to win.
When to use — and when not to
Use Bolt.new when you want a working full-stack app this week, not this month. Perfect for personal tools, internal dashboards, hackathon projects, class projects, and “wouldn't it be cool if X existed” experiments.
Do not use Bolt.new when you're building something with real scale requirements, strong security/compliance requirements (HIPAA, PCI, SOC2), or a long-lived codebase a team will maintain for years. For those, Bolt can still be useful as a scaffolder — but plan to port the code into a real engineering setup afterward.