Best Practices
Expert strategies for writing effective prompts, managing complex projects, understanding limitations, and getting the most from Bolt.new.
Writing Effective Prompts
Be Descriptive
Include specifics about layout, colors, features, and behavior. "A dashboard with a dark sidebar, stats cards at the top, and a data table below" beats "build a dashboard."
Specify Technology
Mention the framework, styling approach, and key libraries upfront. "Using React with Tailwind CSS and Recharts" gives clear direction.
Build Incrementally
Start with core features, then add complexity through follow-up prompts. One or two features per iteration produces the best results.
Reference Existing Code
When iterating, reference specific files or components: "Update the Header component to include a user dropdown menu."
Iterative Development Strategy
- Foundation first: Start with the layout, navigation, and basic structure
- Core features: Add the primary functionality one feature at a time
- Data layer: Connect databases, APIs, and authentication
- Polish: Refine styling, add animations, improve responsiveness
- Edge cases: Handle errors, empty states, loading states
- Deploy: Test in production and iterate on feedback
When to Use Bolt vs Manual Coding
| Use Bolt.new When | Use Manual Coding When |
|---|---|
| Rapid prototyping and MVPs | Complex business logic with many edge cases |
| Landing pages and marketing sites | Large-scale enterprise applications |
| Internal tools and dashboards | Performance-critical applications |
| Hackathon projects | Projects with strict code review requirements |
| Learning new frameworks | Maintaining existing large codebases |
| Client demos and presentations | Apps with complex state management |
Limitations to Know
- Context window: Long conversations may cause the AI to lose track of earlier changes. Start a new conversation for major feature additions if needed.
- Complex state management: Very complex state logic (e.g., Redux with many slices) can be harder for AI to manage correctly.
- Backend limitations: WebContainers run Node.js - other backend languages (Python, Go, Java) aren't supported.
- Large file handling: Very large files or many files can slow down the in-browser runtime.
- Third-party integrations: Some npm packages may not work perfectly in the WebContainers environment.
Cost Management
- Free tier: Limited number of messages and projects per month
- Paid plans: More messages, priority access, and additional features
- Efficient prompting: Write clear, specific prompts to reduce the number of iterations needed
- Export early: Once the basic structure is right, export to GitHub and continue development locally to save messages
Handling Complex Apps
Frequently Asked Questions
Is Bolt.new free?
Bolt.new offers a free tier with limited usage (messages per day and active projects). Paid plans provide more messages, faster generation, and additional features. Check bolt.new for current pricing.
Can I use my own API keys with Bolt.new?
Yes, you can configure environment variables and API keys for third-party services like Supabase, Firebase, Stripe, etc. These are stored securely and used during builds and deployments.
Can I edit the generated code manually?
Yes, Bolt.new includes a full code editor. You can manually edit any file, and your changes will be preserved when the AI makes additional modifications. However, be aware that major AI changes might overwrite your manual edits in affected files.
What happens if Bolt generates buggy code?
You can ask Bolt to fix the issue by describing the bug: "The submit button doesn't work - clicking it should save the form data and show a success message." Bolt will analyze the issue and generate a fix. You can also manually edit the code in the built-in editor.
Can I use Bolt.new for commercial projects?
Yes, you own the code generated by Bolt.new and can use it for commercial projects. However, review the terms of service for the latest details on usage rights and any restrictions.
How does Bolt.new compare to GitHub Copilot?
They serve different purposes. GitHub Copilot assists with code completion within an existing IDE (VS Code, JetBrains). Bolt.new generates entire applications from descriptions. Copilot is for line-by-line coding assistance; Bolt.new is for rapid application scaffolding and prototyping.
Ready to Go Deeper?
Live instructor-led courses from our partners. Affiliate disclosure.
AI & ML Courses - 30% Off
Live instructor-led AI, machine learning, data science, and cloud courses for working professionals. Use code Limited30 at checkout.
EdurekaDataCamp - AI & Data Science
Hands-on Python, machine learning, and AI courses with interactive exercises and real projects.
DataCampedX - Top AI Courses
University-level AI courses from MIT, Harvard, Stanford. Earn certificates that employers recognize.
edX