Getting Started Guide for Bolt.new
This document should serve as a starting point for all users interested in using Bolt to help code up their next project! It is highly recommended that non-developers work through this document from beginning to end before getting started, but feel free to jump around to the topics as you see fit.
This document represents a starting point to help those interested in using Bolt to build their next application! Reviewing this document before you start using Bolt.new will be the most beneficial in helping you get the most out of using Bolt. However, if you’ve started already building, this document should still prove as a helpful guide to get the most out of Bolt!
That being said, many of you are builders and learn best by doing, and if that’s you, please feel free to start building on Bolt.new now!
Do I need to know how to code?
What is Bolt?
Bolt.new is an in-browser AI web development agent that leverages StackBlitz’s WebContainers to allow for full stack application development. The application presents users with a simple, chat-based environment in which one prompts an agent to make code changes that are updated in the real time in the WebContainers dev environment.
Screenshot taken of bolt.new landing page
It’s important to call out that we consider Bolt a tool to help users build full stack web applications. It will require smart prompting by the user and some understanding of web development (either through experience or research) to build more complex applications!
How does it work?
To start building with Bolt, users prompt Bolt by telling it what they want to create. In the example below, you can see that you can start an app with as few as 5 words!
However, users can improve the performance of Bolt as well as get better results by giving Bolt details and specificity on what they are looking to build. In the below example, the user clicked the icon to create an enhanced prompt, but users can feel free to use their own words to provide as much detail as possible.
Prompt quality is one of the best ways to ensure that you are using Bolt most effectively. The more specificity you are able to provide Bolt, the better your results should be!
What is an LLM?
LLM stands for Large Language Model. The way that Bolt is able to generate code for you is our integration with Anthropic’s Claude 3.5 LLM model, which provide high-level programming performance. OpenAI’s ChatGPT is another popular LLM that you may have heard of or used yourself.
Without getting too detailed, LLMs essentially work as a very smart autocomplete. It doesn’t necessarily ‘know’ information, but based on all of the training information it has been fed, it is able to generate ‘new’ thoughts based on patterns that it detected during training.
In the case of Bolt, the LLM is helping you generate code that aligns with your prompt, based on similar applications with similar specifications that it has been trained on.
We’ve actually used an LLM to help come up with the explanation in this section of what an LLM is!
What are tokens?
Understanding tokens and token usage is a critical piece of ensuring that you use Bolt effectively and maximize the benefits of using Bolt for your next project.
A simple definition of tokens:
Tokens are small pieces of text
“I”, “love”, “cats”, “!” are all examples of a token!
LLMs process text as tokens, analyzes them, and then predicts the next token(s) to generate a response.
Token limits: LLMs can only handle a certain number of tokens at a time and the total includes both:
- the input you give (e.g., a long question or document)
- the output it generates (e.g., the response, or the code you get back)
Costs: if you’re using an LLM through a paid service (like you are with Bolt), costs are often calculated based on the number of tokens processed. Fewer tokens = lower cost.
In this table, you can find a rough guide for estimating token costs for code tasks:
As you can imagine, the larger and more complex your app grows, the more expensive (both from a token perspective as well as literal costs) it will become!
Please note, the costs can grow very quickly, so please refer to this helpful learning material to ensure you don’t waste tokens unnecessarily!
How much does it cost?
current as of Jan 7, 2025, may be different upon viewing pricing page today
current as of Jan 7, 2025, may be different upon viewing pricing page today
Important to note regarding token usage. Tokens on a subscription plan do not rollover month to month. If you are on the Pro 100 plan and only use 20 million tokens, upon your next billing cycle, you will go back 55M tokens, NOT 75M tokens.
However if you purchase extra tokens as as token reload, those will rollover into the new month as long as you are on a paid subscription.
Can I really build a full-stack application with Bolt?
Yes! Below you will find examples of many users who were able to successfully build, deploy, and launch using Bolt!
If you go to the Video Tutorials section of our Help Center, you can find many use cases of users that have successfully launched their applications through Bolt!
While we continue to improve deep integration with services for authentication (e.g. OAuth) and database (e.g. Supabase), resolving errors that arise when using these services often requires additional technical research (such as through official documentation or another LLM) into the problem.
Using Bolt “correctly”
Bolt, like any other LLM really boils down to prompting effectively to get Bolt to do what you need / want it to do. Again, Bolt is simply a tool that developers and non-developers alike can use to build fully functioning applications, and learning how to prompt effectively is using your tool most effectively!
We highly recommend that you spend about 1 hour or more reading through the content found here on Prompting Effectively: How to talk to Bolt as well as doing more external research on prompt engineering, a subject that continues to grow as far as tips and tricks daily.
Do I need to know how to code?
The short answer: no!
However, you do need to be very clear with Bolt about:
- what you want Bolt to do
- how it should feel for a user of your web application
- what success looks like
Keeping principles like these in mind when prompting Bolt will increase performance and improve your chances of successfully building through Bolt.
We highly recommend users to do research on LLMs along with technical issues that they run into outside of Bolt as LLMs have the potential to run into error loops and cannot solve all issues on its own.