Build Your Business Website with IndexDock Starter

By the end of this course, you will have a real website for your business, live on the internet, that you own. It will have your business name, your services, your contact details, and it will be built to a professional technical standard, even though you will not write a single line of code yourself.

You do not need any previous experience with computers beyond using a web browser. You do not need to know what GitHub is, what Cloudflare is, or what an AI coding agent is. This course explains every one of those things in plain English, one small step at a time, exactly when you need to know it.

Everything in this course happens in your web browser. You will click buttons, type answers to questions, upload files when asked, and approve things you are happy with. Nothing more technical than that is required of you.

There's a second, less obvious benefit to this course, worth knowing from the very start. The way you'll work through it, describing what you want, reviewing what the agent produces, and asking for changes until it's right, is genuinely how a great many professional developers build software today. AI coding agents like the one you'll use here aren't a simplified, beginner-only version of the real thing; they're a real and increasingly common part of how software actually gets built, by hobbyists and experienced developers alike.

So while the concrete result of this course is your finished website, what you're practicing along the way is a real, transferable skill: how to work alongside an AI coding agent effectively. That's worth having in its own right, well beyond this one project.

There are two participants in building your website: you, and an AI coding agent (an automated assistant that can read and write computer code on your behalf). It helps to understand exactly what each side is responsible for, so you never feel like you are supposed to know something you were never taught.

The agent does all of the technical work. It writes the website's code, tests that the code works, fixes problems it finds, and publishes updates. It never asks you to write code, use a tool called Git, manage anything called a branch, or do anything else technical.

You are responsible for something different but just as important: the truth about your business. You tell the agent what your business does, who your customers are, what makes you different, and what you want visitors to do when they land on your website. You provide any logo or photographs you want used. And at a handful of key moments, you approve the direction before the agent continues. The agent is not allowed to guess or invent facts about your business, so your honest, detailed answers are what make the final website good.

Think of it like hiring a skilled website builder who has never met you: they can build anything, but only if you tell them clearly what you need.

Two things follow from that split. What the agent writes is a draft for you to approve, not a finished professional deliverable: read it before you publish, because it can be wrong. And the website that results is yours — its content, its claims, and whether it meets the rules where you operate are your responsibility, not IndexDock's.

This course, and the free starting kit it uses, is designed for one specific and very common kind of website: an informational or lead-generation website for a freelancer, a local service business, or a solo professional. Lead generation simply means a website whose main job is to make interested visitors get in touch with you, rather than to sell products directly on the site.

A website built this way typically has between one and five main pages, all in a single language. It can explain your business, list your services, show examples of past work, build trust with visitors, and give people simple ways to contact you, such as a phone number, an email address, or a messenger link.

This course is not the right fit for building an online store with a shopping cart and checkout, a website where users create accounts and log in, a booking system built from scratch, or anything that needs to store private customer data in a database. If your idea includes any of those things, don't worry: your AI agent will explain a simpler version of your idea that still works well as this kind of website, or point out that a different kind of tool is needed. Knowing this now just means you won't be surprised later.

Before you continue, one reassurance worth keeping in mind: this course is entirely self-paced. There is no clock running and no deadline to meet. If at any point you feel overwhelmed, confused, or simply tired, it is completely fine to stop, close the tab, and come back later, whether that's in an hour, tomorrow, or next week. Nothing about your progress, your website, or your accounts will be lost while you're away. Take this at whatever speed feels comfortable to you.

GitHub is a website that stores files and keeps a complete history of every change ever made to them. For this course, GitHub will store all of the files that make up your website: its text, its images, its layout, everything.

The place where your website's files live on GitHub is called a repository. You can think of a repository as a single project folder that lives online, which both you and your AI agent can see and work in. You will create your own repository early in this course, and it will belong to you.

GitHub accounts are free to create, and creating one is one of the very first steps in this course.

Writing a website's files is only half the job. Those files also need to be placed somewhere that anyone on the internet can visit and load them in their own browser. That is called hosting, and the act of making a new or updated version of your website available to visitors is called deploying.

Cloudflare is the service this course uses to host your website. Once it is connected to your GitHub repository, Cloudflare automatically takes the latest version of your website's files and makes them available at a web address anyone can open. Every time your AI agent finishes a piece of work and saves it, Cloudflare notices and updates your live website automatically, with nothing for you to click.

Cloudflare accounts are also free to create for what this course needs.

An AI coding agent is a computer program that can have a conversation with you in plain English, and can also read, write, and test computer code by itself. You describe what you want in your own words, and it does the technical work of turning that into a real, working website.

This course uses Claude Code, made by a company called Anthropic. You will interact with it in a browser, at a website called Claude Code on the web, and it is capable of completing this entire course from start to finish.

You will interact with your chosen agent almost entirely through one ongoing conversation, similar to a chat app. You type a message describing what you want, or answering a question it asked you, and it responds and gets to work.

A prompt is simply the message you type to an AI agent. That's the whole idea: it is just written instructions, in your own everyday language, telling the AI what you want it to do or answering a question it has asked you.

You do not need any special skill or secret wording to write a good prompt. The best prompts are simply clear and specific, the same way you would explain something carefully to a new employee who is eager to help but cannot read your mind. If you want a particular thing, say so plainly. If something looks wrong, describe exactly what looks wrong. Later in this course, in the lesson on asking for changes, you will see this idea put into practice in detail.

Throughout this course you will send several prompts: the one that starts the whole project, and others whenever you want to ask a question or request a change.

Claude Code on the web requires a paid Claude plan to complete this project. There is currently no free version that provides enough usage to build a complete website through this course. If you sign up expecting a free option to be enough, you will run out of usage partway through, so it's best to know this up front and go straight to a paid plan.

The good news is that you do not need the most expensive plan available. The entry-level paid Claude plan gives you enough usage to build a complete small-business website through this course. You are welcome to choose a more expensive plan if you want extra usage headroom or access to more capable models, but it is not required to finish this course successfully.

Weighed against what you get in return, a small monthly fee for one or two months in exchange for a finished, professional business website is a strong trade.

It's worth knowing that the money you spend here buys you a general-purpose AI subscription, not a one-time pass that expires once your website is finished. The same Claude subscription keeps working for you afterward, for all sorts of everyday things that have nothing to do with building websites: analyzing or describing photos, looking up and explaining information, generating images, and countless other everyday tasks people use these tools for.

That also means the same subscription is ready whenever you want to build another website. If you start a second business, or simply want to try again with a different design or a different set of business details, you can create a new repository from the IndexDock Starter template and walk through this same course again, using the subscription you already have. Nothing about your paid plan ties you to building only one website.

You may have heard the phrase "AI visibility" or wondered whether a website can be found not just by search engines like Google, but also by AI assistants that people ask questions to. This starter kit is built with that concern in mind, and it's worth understanding exactly what that promise does and does not mean, in plain terms.

What this course's website gives you is technical readiness. That means your website is built so that computer programs, including search engines and AI assistants, can correctly read your pages, understand what your business does, and describe it accurately if someone asks. Nothing about your website will secretly block or confuse those programs.

What this course does not promise is guaranteed rankings, guaranteed traffic, or a guarantee that an AI assistant will mention your business by name. Nobody can honestly promise those outcomes, because they depend on many things outside any single website, such as competition, location, and time. Be cautious of anyone who promises otherwise. What you are getting is a website with no technical barriers standing in the way of being found and understood correctly.

A few short definitions are useful here, only because they explain why this matters, not because you need to use these words yourself.

Crawlable means that a computer program can successfully open and read your website's pages, the same way you can visit them in your browser. Some websites are built in ways that accidentally hide their content from these programs; the websites this course builds are not.

Structured information means that the important facts about your business, such as its name, its services, and its location, are labeled clearly in the website's underlying code, not just visible as a picture or scattered plain text. This labeling helps computer programs understand what a piece of text actually means, rather than just seeing letters on a page.

Accessible, in this context, means your website works well for as many visitors and devices as possible, including people using screen readers or other assistive technology, and people using different browsers.

Together, these three things are what "ready for AI search" means in this course: a website with nothing standing in the way of being read, understood, and represented correctly, by humans and by computer programs alike.

Indexing is what search engines and AI assistants do behind the scenes: they visit a website, read its pages, and add what they learn to their own giant catalog, so that later, when someone searches for something related, that website can show up as a result. A page has to be indexed before anyone can ever find it that way; being indexed is what makes a website findable at all.

While you're still building, your website carries a quiet instruction that tells search engines and AI assistants not to index it yet, so nobody accidentally stumbles onto an unfinished site while you're still working on it. This is completely normal and expected; it isn't a sign that anything is wrong or that the technical readiness described above isn't working.

That instruction comes off automatically once you give your final approval on the completed website, later in this course. From that point on, your site becomes genuinely discoverable, the way everything in this module has described.

Open your web browser and go to [github.com](https://github.com). Look for a button to sign up or create an account. You will be asked to choose a username, enter an email address, and create a password. Follow GitHub's on-screen instructions, which usually include confirming your email address by clicking a link GitHub sends you. Once your email is confirmed, your account is ready to use.

Keep your username, email, and password somewhere safe. You will need to be signed in to GitHub several more times during this course.

Open your web browser and go to [cloudflare.com](https://cloudflare.com). Look for a button to sign up. You will be asked for an email address and a password. Follow Cloudflare's on-screen instructions, which will also include confirming your email address. Once confirmed, your Cloudflare account is ready.

You do not need to buy anything or enter payment details to create a Cloudflare account for what this course needs.

Now is the point where you create your account with Anthropic, the company behind Claude Code.

Go to [claude.ai](https://claude.ai) and sign up for an account, following the on-screen instructions to confirm your email address. Afterward, upgrade to a paid Claude plan, since Claude Code on the web requires one, as explained earlier in this course.

Take this step seriously even if you have never used an AI chat tool before; the sign-up process itself is short and guided, and this course will tell you exactly what to do with the account once it exists.

A template is a ready-made starting project that you can copy into your very own new repository with one click, instead of building something from nothing. IndexDock Starter is a template: it already contains a small, working starter website and everything your AI agent needs to build your real website on top of it. Using a template means you begin with something that already works, rather than an empty folder.

Open your browser and go to [github.com/IndexDock/website-from-scratch](https://github.com/IndexDock/website-from-scratch). Make sure you are signed in to your own GitHub account first.

On that page, find and click the button labeled "Use this template", and then choose "Create a new repository" from the option it shows you. GitHub will ask you to give your new repository a name; choose something related to your business, such as its name. When GitHub asks whether the repository should be Public or Private, choose Private. Your website will still go live for everyone to visit; this setting only controls who can read your files on GitHub. Then click the button to create the repository.

Once it is created, leave it exactly as it is. Apart from uploading your logo and photos later, when the agent asks, you never open or edit a file yourself. Your new repository is now your own copy of the starter kit, owned by you, ready for the next steps in this course.

A Worker, in Cloudflare's terms, is the piece of Cloudflare's system that actually serves your website's pages to visitors. When someone types your website's address into their browser, it is Cloudflare's Worker that responds with your website's content. You do not need to understand how it works internally; you only need to create one and connect it to your GitHub repository, which the next lesson walks through one click at a time.

Make sure you are signed in to your Cloudflare account, then follow these steps exactly, in order. Two of them ask you to type a short command into a box. You are copying two lines into a form, not writing code: the course gives you the exact text, and you never need to understand what it does.

1. In the left sidebar, click Build.
2. Under Build, click Compute (Workers & Pages).
3. Click the blue button labeled "Create application".
4. Click "Continue with GitHub".
5. Select the GitHub repository you created from the IndexDock Starter template.
6. When asked for a build command, enter npm run build.
7. When asked for a deploy command, enter npx wrangler deploy.
8. If you see a setting called "Builds for non-production branches", leave it switched off. This setting is not needed for this course, and turning it on will use up Cloudflare's free build allowance faster than necessary.
9. Click the Deploy button.

If, during this process, Cloudflare asks you to set up your account's workers.dev subdomain, do that now. It is a one-time setup step that only appears the first time, so if you see it, complete it before continuing.

Cloudflare will now build and deploy your starter website. This usually takes a minute or two.

After your Worker is created, Cloudflare needs one more setting turned on before your website is reachable at its web address. Follow these steps exactly.

1. In the left sidebar, click Build.
2. Under Build, click Compute (Workers & Pages).
3. Click on the Worker you just created to open it.
4. Click the Domains tab.
5. Find the entry labeled "production" and turn it on.

Once this is switched on, your starter website becomes reachable at its workers.dev web address.

In the Cloudflare dashboard, on your Worker's page, click the "Visit" button in the top-right corner to open your website's workers.dev web address in a new browser tab. You should see the IndexDock Starter's starter page, confirming that your website is online. This is not your finished business website yet; it is a temporary starting page that confirms everything is connected correctly. Your real business website will appear at this exact same web address as the agent builds it later in this course.

If an error appears instead of your website when you open the address, don't worry. This usually just means Cloudflare is still finishing the deployment behind the scenes. Wait roughly one minute, then reload the page, and you should see the actual starter website appear.

Take a moment to enjoy this: your website's address now works absolutely everywhere on the internet. Open it on your computer, open it on a tablet, open it on your phone, it will work the same way on all of them. You can even send the link to a friend or family member right now and let them be among the first to see it. It's a small milestone, but a real one, and there's a lot more still to come.

If you look near the bottom of the page, you'll notice a small, subtle line reading "Built with IndexDock Starter". This is expected, not a mistake. It's a quiet credit to the free starter kit that this whole course is built around, and it will stay on your website even after your real business website replaces this starter page. Keeping it is the default, not a condition: if you would rather not have it, ask your agent to remove it.

On GitHub, a star is a simple way of showing appreciation for a project, similar to a "like" on other websites. When you star a repository, it becomes more visible to other people browsing GitHub, and it lets the people who built the free starter kit you are using know that it is genuinely helping someone.

You have now successfully deployed your own copy of the starter kit, so this is a fitting moment to give it a star if you'd like to.

To do this, open your browser and go to [github.com/IndexDock/website-from-scratch](https://github.com/IndexDock/website-from-scratch), the original template page, not your own copy of it. Near the top of the page you will see a button labeled "Star". Click it.

Giving a star is entirely optional and never required to continue this course. Nothing about your website or your progress depends on it.

You have just done something real: you created a GitHub account, a Cloudflare account, your own repository, and a live website address, and you saw an actual page load in your browser. That is proof this whole process genuinely works.

Now, before you open your AI agent and start the actual work of building your business website, read the rest of this course all the way through, from here to the end. It is not long, and none of it requires you to do anything yet. Its purpose is to prepare you: it explains what your AI agent will ask you, what each stage of the process looks like, and a handful of important details, such as how to upload your logo correctly and how to ask for changes effectively, that are much easier to get right the first time than to fix afterward.

Once you have read through to the end of this course, come back to the module named "Starting the Agent and the Journey Ahead" to begin the real work.

Claude Code needs permission to see and work inside your GitHub repository before it can build anything. This is done through an authorization step, where you tell GitHub that Claude Code is allowed to access your repositories. This step happens once, in your browser, and takes only a minute.

Open your browser and go to [claude.ai/code](https://claude.ai/code) and sign in. If you haven't connected GitHub yet, Claude Code will prompt you to do so; follow that prompt to install the Claude GitHub App and grant it access to your repositories. If you're ever given a choice between granting access to all repositories or only some, choosing all repositories is the simplest option and avoids any chance of confusion later.

In Claude Code on the web, you choose which underlying AI model does the work yourself, from a small menu or dropdown that sits near the message box in your conversation. Look for it before you send your very first message, since it is easiest to set correctly from the start.

That menu lists several models, one below the other, ordered from the most powerful at the top to the more modest ones further down. A more powerful model tends to produce more thoughtful, higher-quality work, but it also uses up more of your usage allowance for the same amount of work. A model further down the list is less powerful, but it goes easier on your usage allowance, so you can do more with it before running into the limits explained a couple of lessons from now. Picking a model is therefore its own small trade-off between quality and how far your usage stretches, alongside the reasoning-effort trade-off covered in the next lesson.

Alongside the model choice, you will also see an option for reasoning effort, sometimes labeled low, medium, or high. This controls how much time and thought the AI puts into each step of the work. Higher reasoning effort tends to produce a more polished, more carefully considered result, but it takes longer to run.

This matters because of a usage limit explained in the next lesson: a lower reasoning effort, such as low, or medium in some cases, is more likely to finish a complete website within a single five-hour working session. A higher reasoning effort will very likely take longer than five hours to finish in one sitting, meaning your session will pause partway through. That is not a failure; it simply means you continue in a later session, as the next lesson explains, and the end result is generally a better-quality website.

There is no single correct choice here. If you would rather have a finished website sooner, even if slightly simpler, choose a lower reasoning effort. If you are comfortable coming back more than once in exchange for a more polished result, choose a higher one. It is entirely your decision, and either path finishes successfully.

Claude Code has a usage limit that resets every five hours. If you reach it while your agent is working, your agent will simply stop and tell you it needs to wait. When five hours have passed since your session began, the limit refreshes, and you can continue by typing a short message such as "Continue working". Your agent will pick up exactly where it left off, because its progress is saved, as explained further in a later module.

There is also a separate weekly usage limit. For a project the size of the website you are building in this course, that weekly allowance is comfortably enough to finish in a single week, even accounting for the five-hour pauses described above.

Whatever happens, remember this: your progress is always saved. Coming back after a short break, an overnight pause, or even a much longer gap of weeks or months costs you nothing. Your agent picks up exactly where you left it.

Claude Code offers a choice in how much it checks with you before making changes, called Accept Edits or Auto mode. In the more cautious modes, the agent stops and asks your permission before every single change it wants to make, which means you need to stay at your screen approving one small step after another. In Auto mode, the agent keeps working through many steps on its own without needing your approval each time.

For this course, Auto mode is generally recommended. The website you are building is a lightweight application that does not request access to sensitive parts of your computer, so the risk of letting the agent proceed without constant approval is low, and it means you can go do something else while it works instead of sitting and clicking approve over and over.

One important clarification: Auto mode is not the same as fully unsupervised. You will still need to come back and check on your agent's progress from time to time, especially whenever it sends you a notification, since it will still pause at genuine decision points, such as approving a design or reviewing the finished website, that only you can make.

While typing in Claude Code, you will sometimes notice a greyed-out suggestion already sitting in the message box, showing you a relevant message you could send next. When you see this, you can press the Tab key on your keyboard to instantly fill that suggested message into the box, rather than typing it yourself. This is a small convenience worth knowing about, since it will come up throughout your work with the agent.

Open Claude Code on the web and open the repository you created earlier from the IndexDock Starter template. With your model and mode chosen, as covered in the previous module, type or paste this exact message to begin:

Start or continue my IndexDock Starter website. Read the repository state first, tell me where we are, and guide me through the next approved step.

Send that message, and your agent will take it from there, starting with a short round of questions.

Building your website happens in a series of stages, and it helps to know what they are before you're in the middle of them. Here is the journey in plain terms.

• A short setup stage, where the agent checks that everything is connected correctly.
• A short scope stage, where the agent asks a couple of quick questions to confirm your idea fits what this Starter can build, and, if needed, suggests a simpler version that does.
• A stage for understanding your business, where the agent asks you questions about what you do, who your customers are, and what you want the website to achieve.
• A design stage, where you are shown a few different visual directions for your website and choose the one you like best.
• A building stage, where the agent writes all of the actual website code based on everything you approved.
• A quality-checking stage, where the agent automatically tests the website it built and fixes anything it finds.
• A publishing stage, where your finished, approved website becomes the real, indexable version the public sees.

Each of these stages is covered in its own module later in this course, in the order they happen, so you will have a fuller explanation of each one right before you reach it.

Once the setup stage finishes, your agent will first ask you a couple of quick scope questions, just to confirm your idea fits what this Starter can build. If your idea includes something outside that scope, the agent will suggest a simpler version that still works well, rather than turning you away.

With that quick check done, your agent will start asking you the real questions about your business: things like what you do, who your ideal customers are, what problems you solve for them, what services you offer, what makes you different from competitors, what you want visitors to do on your website, and whether you have any existing photos, logos, or other materials.

The agent asks all of this because it is not allowed to guess or invent facts about your business. Every claim on your finished website needs to come from something you actually told it. That is a deliberate safeguard: it means your website will never show a made-up statistic, a fake customer review, or a promise you never actually made.

When answering these questions, resist the urge to keep your answers short or vague. The more real, specific detail you give the agent about your business, the better it can structure your website: which pages to create, what to say on each one, and which details to highlight. Vague answers lead to a generic website; detailed answers lead to one that actually sounds like your business.

There is no such thing as too much detail here. If something feels like a small or obvious fact about your business, mention it anyway. You are the only source of truth the agent has.

Once your business questions are answered, the agent will put together a package summarizing your business, the list of pages your website will have, and what each page will contain. Read through this carefully. Nothing moves forward to the design stage until you say yes. If anything looks wrong, missing, or not quite right, this is exactly the moment to say so, as covered in more detail in the module on asking for changes later in this course.

This is one of the most important technical details in this whole course, so read it carefully. If you attach a logo or photo directly into your conversation with the agent, the agent can look at the image and describe what it sees, but it cannot save that image's actual file into your website. Attaching a picture to the chat is not the same as delivering it. If you rely only on attaching images in the chat, your finished website will not actually contain them, even though the agent could technically see them.

The only way to get your real logo and photo files into your website is the upload method described in the next lesson.

When your agent needs your logo or photos, it will give you a direct web link that opens GitHub's upload page for your specific repository. Click that link. It will take you straight to a page ready for you to add files, with nothing else to find or choose.

On that page, drag your logo or photo files into the upload area, or use the button to browse and select them from your computer. Once your files are added, click the button GitHub already has selected to confirm the upload. You do not need to choose a folder, type a file path, or pick a branch; none of that is necessary. Once you have uploaded the files this way, tell your agent you have done so, and it will find them and put each one exactly where it belongs in your website.

Don't worry if you don't have a logo. In that case, the agent will propose a simple, honest placeholder, such as your business name styled with a chosen font, or its first letter or initials in a simple mark, and show it to you for approval before using it anywhere. You can always replace it with a real logo later, at any point in this course or afterward.

Once your business details are approved, the agent will create a small number of different visual directions for your homepage, usually three, or two when your examples point strongly one way, so you can compare real options rather than imagining one from a description. Each one will use your real approved business content, just presented differently: different fonts, colors, layouts, and overall feel. You will look at these and choose the direction that fits your business best, or ask for changes.

Before this stage, collect screenshots of websites whose design you genuinely like, for any kind of business, not necessarily one similar to yours. Your agent will ask for eight to fifteen of them in a single request, and more are welcome. They travel through the same GitHub upload link as your logo and photos, so it saves time to capture them in advance. Having real examples ready makes an enormous difference to how well your finished design matches what you actually pictured.

It helps even more if you can say specifically what you like about each example: is it the font, the colors, whether buttons have sharp or rounded corners, how much empty space is used, how big the images are, or something else entirely? The same goes for anything you know you dislike in general, in any website you've seen. You do not need to be a designer or use correct design terminology to explain this. Any honest description in your own words, even an imperfect one, gives your agent something real to work from, and is far more useful than no preference stated at all.

When your design directions are ready, the agent will show them to you on a live, working preview so you can judge them properly rather than looking at a flat picture. Look closely at each one. You have a few options: approve the one you like, ask for a revised version of one you almost like, reject all of them and ask for a different direction entirely, or mix and match, for example telling the agent you want one concept as your base but with a specific element borrowed from another, such as its color scheme or its layout. Nothing about the design is final until you approve it, so take your time here.

Once you approve a design direction, the agent moves into building the real website. In plain terms, it is writing the website's underlying code, section by section, based on your approved business content and your chosen design. Along the way, it also runs its own basic checks on what it just wrote and fixes anything obviously broken before moving on.

Beyond that first pass, a second, more thorough round of checks runs automatically behind the scenes after the agent saves its work, covering things like whether the website looks right on phones as well as computers, and whether it meets accessibility standards so people using assistive technology can use it too. The agent reads the results of these checks and fixes anything they turn up; you don't need to run or watch this step yourself.

You do not need to understand any of the technical detail behind this. It is simply useful to know that "building" is not a single instant action; it happens in layers, with checks happening both as the agent writes and again automatically afterward.

Throughout the building stage, your live website, at the same web address you confirmed earlier in this course, updates itself automatically as the agent completes pieces of work. Whenever your agent tells you it is a good time to look, open that address again in your browser and review what has changed. You never need to do anything to make an update appear; it happens on its own once the agent has saved its work.

Here is something worth saying plainly: you are allowed, and encouraged, to ask for changes at any point in this process, whether you are reviewing your business summary, your design, or your finished website. You do not need to feel like you're being difficult or ungrateful by asking for something to be different. Speaking up clearly about what you want changed, at every single stage, is not a side effect of using these tools well; it is the actual skill involved in working with an AI agent effectively. People who get the best results from AI agents are the ones who look closely and ask for exactly what they want, as many times as it takes.

A simple pattern makes asking for changes effective. First, look at whatever the agent has just shown you: a summary, a design, or the live website. Second, inspect it closely rather than glancing at it; read the actual text, look at spacing and colors, click around if it's the live site. Third, identify specifically what is wrong, missing, or could be better. Fourth, ask for that exact change in plain language. Repeat this loop as many times as you need to, at any stage, until you are genuinely happy with the result.

If something on your website looks wrong in a way that is easier to show than to describe, take a screenshot of it and add it to your message to the agent, along with a plain English description of what should be different, for example pointing out which part of the picture you mean and what you'd like to see instead. A screenshot paired with a short explanation is often faster and clearer than trying to describe a visual problem in words alone.

• On Windows, press the Windows key, Shift, and S at the same time to open the snipping tool, then drag over the area you want to capture.
• On a Mac, press Shift, Command, and 4 at the same time, then drag over the area you want to capture.

Since many visitors will view your website on a phone, it is worth checking it there yourself, not just on the computer you have been working on. Send your website's live web address to yourself in some way that is easy to open on your phone, such as a message to yourself in a messaging app or an email you send to your own address. Then open that link on your phone's own web browser and look through your website the way a real visitor would: check that the text is readable, that buttons are easy to tap, and that nothing looks squeezed or cut off. If you spot a problem, use the same change-request approach described earlier in this module to have it fixed.

If what you spot is a visual problem worth capturing, take a screenshot right there on your phone, the same way you normally would. Then open [claude.ai/code](https://claude.ai/code) in your phone's browser, select the chat where your website's work is happening, and attach the screenshot to your message there, along with a plain English description of what should be different.

Right now, your website's address ends in workers.dev, which is a free address Cloudflare gives every project. A domain is a web address you choose and register specifically for your business, such as yourbusinessname.com, which looks more professional and is easier for customers to remember.

Once your website is finished and you're happy with it, your agent will walk you through connecting a real domain if you want one. It will show you the exact domain name available and the exact price before anything happens, and it will wait for your clear approval before registering or purchasing anything at all. Nothing is ever bought automatically or without you explicitly agreeing to the exact domain and price shown to you.

If you decide you don't want a custom domain right now, that's completely fine. Your website keeps working perfectly well at its workers.dev address, and you can always add a real domain later.

Once your website is published, it isn't set in stone. You can come back at any time, whether that's next week or next year, open the same task with your agent, describe what you'd like changed, and the agent takes care of it the same way it built the site in the first place. There's no separate setup to redo and nothing extra to configure; you simply describe the change.

Not every change goes through the exact same steps, but you never need to figure out which kind yours is; the agent handles that automatically. If you ask for a real change of design direction, the agent shows you new options and waits for your approval the same way it did the first time. If you ask for another kind of visible change, such as new wording, a new page, or an updated photo, the agent makes the change, quietly re-runs its usual checks behind the scenes, and shows you the result to confirm before it goes live. If the change is something small and purely behind the scenes that no visitor would ever notice, it can simply go live without needing a fresh approval from you. Either way, your live website only ever changes to something you've seen and are happy with.

Once the agent believes your website is complete, it will ask you to review the entire finished site one more time before it becomes the public, real version of your website. Look through it carefully, the same way described in the module on asking for changes, and ask for any final adjustments you want before giving your approval. This is your last chance to request changes before publication.

You created your repository as private right at the start, so there is most likely nothing to change here. This is a final check. Your live website is public and stays that way; your files on GitHub are what stays private.

Open your repository on GitHub and look next to its name at the top of the page. It should show a small label reading Private. If it reads Public instead, you picked the other option when creating it — switch it now:

1. Click Settings.
2. Stay on the General tab.
3. Scroll all the way to the bottom, to the section called Danger Zone.
4. Click Change visibility.
5. Choose Make private, and confirm.

Either way your live website is unaffected. Making a repository private changes who can read your files, never who can visit your site.

This template and this course are given to you as they are, with no guarantee about the result. It is worth being plain about where that line falls, so that nothing important is left implied.

• An AI agent writes the site, and it can be wrong. Read what it produces before you publish. Treat it as a draft you approve, not as finished professional work.
• No rankings, traffic or business results are promised. The site is built so that search engines and AI assistants can read it properly. Whether they rank it, quote it, or send you customers is not in anyone's control.
• The services used here belong to other companies. GitHub, Cloudflare and Anthropic set their own terms, prices and availability, and can change them at any time.
• Nothing in this course is legal, tax or professional advice. A business website often needs its own privacy policy, imprint or cookie notice, depending on where you and your customers are. That part is yours to get right.

None of this is a reason to hesitate. It is the same split this course has described from the start: the technical work is handled for you, and the truth about your business stays with you.

There is no exam or assignment to submit at the end of this course. You are finished when your website is live and your repository has been switched to private, as covered in the previous lesson. That's it: you now own a real, working business website that you built without writing a single line of code yourself.

If, later on, you want to go further with things like deeper strategy, competitive research, or ongoing support for your website, IndexDock also offers professional services for that. But that is entirely optional, and this free course and the website you built with it stand complete on their own.