Cindy Zhu.
← all free guides
Building & apps No-code build

Build websites worth thousands with Claude and Emergent

Hey, it's Cindy ๐ŸŒฑ A clean small-business website sells for thousands but costs about $20/mo of tools to build. This is the exact Claude + Emergent workflow, from spec prompt to live site to first sale.
the opportunity

The arbitrage, in one line ๐Ÿ’ก

Worth $2,000โ€“10,000What a small business pays for a clean brochure or landing site (Forbes Advisor).
27% have noneMore than a quarter of US small businesses still have no website at all (survey of 1,003 owners).

The old cost was a designer plus a developer. The new cost is one AI tool plan. You buy the tools once, produce a site worth thousands, and keep the difference. It is not a loophole, just the price of building collapsing faster than the price businesses pay.


the kit

What you need before you start ๐Ÿ“ฆ

A Claude accountclaude.ai. The free plan handles the whole strategy and copy step.
An Emergent accountThe free tier (10 credits) builds a preview only. To keep a site live you need Emergent Standard (~$20/mo, 100 credits), because a deployed site uses 50 credits every month.
A business to build forYour own, or a local one: a cafe, a plumber, a dog groomer, a physio, anyone with no site or an outdated one.
Time and moneyAbout 10 minutes with Claude for the plan and copy, under an hour for the first build, and ~$20/mo of tools total.

How the two tools split the work ๐Ÿงฉ

Claude = strategist and copywriter

You describe the business. Claude plans the pages, writes the actual copy, and hands you a build-ready spec. Runs on the free plan.

Emergent = the builder

Emergent builds with Fable 5 (Anthropic's newest Claude-family model), turning that spec into a working full-stack site on a live URL with SSL. Nothing to install.

Why the split matters: Emergent builds far better when the first prompt already arrives organized into pages, copy, and a data model for the lead form. A vague "make me a website for a cafe" wastes credits and produces filler. The prompt below hands the builder exactly that, with the real copy already written.

copy this

Step 1: plan and write it with Claude ๐Ÿง 

Open a new Claude chat and paste the prompt below. Claude asks a few discovery questions about the business, then returns a full multi-page plan, the real copy for every page, and a build-ready spec you paste straight into Emergent. This is the resource you were sent. Copy the whole block.

๐Ÿง  Turns a business into a build-ready spec
You are my web strategist and copywriter. I am building a clean, professional website for a small business, and I will build it in Emergent, an AI app builder that turns a written spec into a live full-stack website. Your job is to plan the site, write all the real copy, and produce ONE build-ready spec I can paste into Emergent as my very first prompt with zero edits.

STEP 1, DISCOVERY. Before writing anything, ask me these questions in a single message, then stop and wait for my answers:
1. What is the business, and what does it sell or do?
2. Who is the ideal customer, and what do they care about most?
3. What are the 3 to 5 core services or offers, in the business's own words?
4. What makes this business the right choice over competitors (real reasons, not fluff)?
5. What is the ONE action a visitor should take (call, book, fill the lead form, visit in person), and what details must the lead form collect?
6. Business name, location or service area, and any phone, email, or hours I should show. If you don't have a logo or brand colours, say so and I will pick a clean default.
Ask only these. Do not write any copy or spec yet.

STEP 2, THE SITE PLAN AND COPY. After I answer, produce the following, in this order:

A. SITE MAP. List the pages and the one job each page does. Default to exactly these five: Home, About, Services, Contact (with a lead form), plus a short Testimonials or Why-us section that can live on the Home page. Do not add pages the business doesn't need.

B. THE ACTUAL COPY, page by page, ready to ship. Write in the business's real voice, warm and clear, benefit-led, no jargon and no filler. For each page give:
   - Home: a headline (the promise in under 10 words), a one-line subhead, 3 short benefit blocks, one social-proof line, and one clear call to action.
   - About: 2 short paragraphs on who they are and why they can be trusted, plus 3 quick credibility points (years, area served, specialisation).
   - Services: for each core service, a short heading and 2 to 3 sentences on what it is and who it's for. End with a call to action.
   - Contact: one friendly line inviting the visitor to reach out, the contact details, and the lead form fields.
   Keep every headline specific to THIS business. No "Welcome to our website", no "We provide quality service". Name the outcome the customer actually wants.

C. THE BUILD-READY SPEC for Emergent. Write it in second person to the builder ("Build a website that...") so it pastes clean. Use exactly this structure:
   - Section 1, What this is: 2 to 3 sentences. A public marketing website for [business], with the pages from the site map. One private inbox screen for the owner to read lead-form submissions. No other logins.
   - Section 2, Pages and sections: for each page, list its sections in order and drop in the real copy from part B verbatim, so the builder places actual words, not placeholders.
   - Section 3, The lead form: every field with its type (text, email, phone, dropdown with the exact options, long text), which are required, and what happens on submit (save to a database, show a friendly on-screen thank-you, and appear in the owner's inbox screen). No emails are sent by the site itself; the owner reads submissions in the inbox.
   - Section 4, Design direction: clean, modern, fast, mobile-first. State the palette (use the business's colours if given, otherwise a warm neutral base with one confident accent), one heading font and one body font, generous spacing, rounded corners, real-feeling section imagery described in words so the builder uses fitting placeholders. Navigation sticky at top, clear call-to-action button in the header.
   - Section 5, Must-haves: mobile responsive, fast loading, click-to-call and click-to-email links, a footer with business name, area served, contact details and hours, and basic on-page SEO (a real page title and meta description per page, written from the copy).
   - Section 6, OUT OF SCOPE (caps): no user accounts for visitors, no payments, no booking calendar, no blog unless I asked, no analytics dashboards. Keep v1 a clean marketing site plus a lead inbox.
   - Section 7, Acceptance tests: 6 to 8 "I can..." checks that prove it works (for example "I can submit the lead form on my phone and see it appear in the owner inbox").
   - End the spec with: "Seed the lead inbox with 3 realistic demo submissions so the screen renders with content on first load."

HARD RULES: no vague words (quality, robust, seamless, world-class); replace them with specifics. Every form field gets a type. Every headline names a real outcome. If my answers leave a gap, make the smallest sensible choice and tag it "ASSUMPTION:" inline so I can override it. Do not invent services, testimonials, or credentials I did not give you; if you need a placeholder testimonial, clearly mark it "[REPLACE with a real client quote]". After the spec, list every assumption in one short block so I can correct it before I paste.

Step 2: build it in Emergent ๐Ÿ› 

Paste the spec and a working preview usually appears in a few minutes, with a full first build done in about 10 to 20. Deploy and you get a live URL with SSL. Try it here: Emergent.

  1. Paste the whole spec first Drop the entire Claude spec in as your very first message. Do not summarize it; the copy and structure are the point.
  2. Answer any questions in one go If the builder asks for clarifications, reply to all of them in a single message.
  3. Preview on phone and laptop Click through every page like a customer would before you deploy.
  4. Deploy, then connect GitHub Deploying gives you the live URL and SSL. Connect GitHub on day one so your code syncs there: it is your safety net, and your proof the site can be handed over cleanly when you sell.
Heads up on credits: a deployed site holds 50 credits/mo (half your Standard plan), so ship once v1 actually looks right, not before.

Step 3: iterate, one change at a time ๐Ÿ”

First builds come out about 80% right, which is normal, not failure. Plan on 2 to 4 quick rounds, and fix in order: structure first, then copy, then the style pass, because a paint pass before a structural change gets clobbered. Everything in [brackets] is a placeholder with a worked example, so swap in your own values.

Three habits that save credits: one change per message (bundle only closely related tweaks); be surgical, not vague ("stack the services cards into one column under 600px" beats "the services page looks off"); and roll back instead of arguing, because rollbacks are free and fix-loops cost credits.

1. Fix the page structure

๐Ÿ—บ Pages and nav before anything else
Adjust the site structure only. The pages should be exactly: [Home, About, Services, Contact]. Move the [Testimonials] block onto the Home page under the benefits section. Put a clear call-to-action button ([Get a free quote]) in the sticky header on every page, linking to the Contact form. Do not change any copy or styling yet. Done when the nav shows exactly these pages and the header call-to-action appears on all of them.

2. Tighten the copy

โœ๏ธ Make it sound like a real local business
Copy-only pass on the [Home] page. Replace the current headline with: "[real headline]". Rewrite the three benefit blocks to lead with the customer outcome, one sentence each, in a warm, plain voice, no jargon. Keep everything else on the page in place. Done when the new headline and three benefits read like a real local business wrote them.

3. Wire up the lead form the way the owner wants

๐Ÿ“ฅ The part that actually captures leads
Adjust the Contact lead form only. Fields: name (required), email (required), phone (optional), "[what do you need help with?]" (dropdown: [Service A, Service B, Service C, Something else]), and message (long text, required). On submit: save to the database, show a friendly on-screen thank-you with the business's phone number, and make the submission appear in the owner inbox screen with a timestamp. The site sends no emails itself. Do not touch other pages. Done when I can submit the form on my phone and see it land in the owner inbox.

4. Add a simple appointment booker (the business-owner move)

Run this after v1 ships. It turns the site from a brochure into something that books clients while you work, with no payments and no calendar integrations, so it stays inside what this stack does well.

๐Ÿ“… Brochure site becomes a booking site
Add a simple appointment booker to the site. New public page "Book a time", linked from the header call-to-action. Fields: name (required), email (required), phone (optional), service (dropdown: [Service A, Service B, Service C]), and a date-and-time picker offering [30-minute] slots inside my business hours ([Mon-Fri 9am-5pm, my timezone]). Block out slots that are already booked so double-bookings are impossible. On submit: save the booking, show a friendly confirmation with the business's phone number, and list it on a private "Bookings" screen for the owner (next to the lead inbox), ordered by date, with a cancel button per booking. No payments and no external calendar sync in v1; bookings live in the app and I check them there. Do not change any other page. Done when I can book a slot on my phone, see it appear on the owner's Bookings screen, and a second visitor cannot book the same slot.

5. Style pass (swap in the business's colours)

๐ŸŽจ Paint it last, on purpose
Style-only pass. Do not change any copy, pages, or the lead form. Apply this look everywhere: background [warm off-white], primary buttons and links [one confident accent, e.g. a deep teal], dark warm-gray body text, one heading font [a clean modern serif] and one body font [a friendly sans], generous spacing, rounded corners, subtle section dividers. Make it feel premium and trustworthy on mobile first. Keep every element where it is; change only colour, type, and spacing. Done when every page uses this palette and still passes the acceptance tests.

6. Bug fix template (reuse for anything broken)

๐Ÿ› Keep the fix scoped to the bug
Bug report. On the [page] page I did [exact action] on [phone or laptop]. Expected: [X]. Got: [Y]. Fix only this bug, do not refactor or change any other feature while you are in there. Tell me the cause in one sentence, then confirm the fix. Done when [the expected behaviour happens].

the money part

Step 4: how to actually sell it ๐Ÿ’ธ

The build is the easy half. Selling is the half that pays. Here is the whole loop.

  1. Find businesses that visibly need it Walk your suburb or search Google Maps for a category (plumbers, cafes, cleaners, groomers, physios). Two kinds convert: those with no site at all, and those with a slow, ugly, non-mobile site. The second group is often easier, because they already believe a site matters, they just have a bad one.
  2. Build the demo first, then reach out This is the unfair advantage. Spend one build on a real site for them (their name, services, a good guess at the copy), deploy it to a temporary Emergent URL, and lead with the link. "I built a new website for [business], here is the live link, tell me what you think" beats any cold pitch, because they can see it, not imagine it.
  3. Make a simple offer "I build clean, fast, mobile sites for local businesses. I already made one for you. If you like it, it's a [one-time fee] to make it yours on your own domain, and [small monthly or nothing] to keep it running." You are selling a finished thing they can see, not a project.
  4. Price for the value, not your cost A clean site is worth $2โ€“10k, so a few hundred to a few thousand is not underpricing. Sensible bands: a few hundred for a fast one-pager, $800โ€“2,000 for a full multi-page site, more if they want ongoing edits. Split it if you like: a one-time build fee plus a small monthly for hosting and changes, which covers your $20 tool cost many times over and makes the income recurring.
  5. Deliver and hand off the domain Once paid, connect their own domain (a free DNS record in Emergent) and give them access, or maintain it for the monthly. If they want full ownership, the code lives in the GitHub repo you connected, so it transfers cleanly. Say up front which one "theirs" means.
  6. Turn one into many The moment you have one live client site, screenshot it, put it in a simple portfolio, and use it to win the next. The first sale is the hard one. After that you are showing real work, and the same build takes you an hour.
Don't oversell it. Charging a few hundred to a few thousand for an hour of tool time is completely fair, because you are selling the outcome (a business that looks credible and captures leads), not your hours. A simple appointment booker is fair game. Do not promise payments, calendar sync, or a complex web app on this stack, and never tell a client it is a bespoke $50k build. Sell a clean, fast, honest marketing site, which is exactly what most of them need.

the handoff

Step 5: connect a custom domain (free) ๐ŸŒ

A live Emergent URL is fine for the demo, but a real client wants their own domain. It is a free DNS record, not an extra bill (they just pay for the domain name itself, usually $10โ€“20/year at their registrar).

  1. Open the site's settings in Emergent On the deployed site, choose to connect a custom domain.
  2. Add the DNS record at the registrar Paste the record Emergent gives you into the client's domain host (GoDaddy, Namecheap, Cloudflare, wherever the domain lives).
  3. Wait for it to verify SSL is issued automatically, so the site loads on https with the padlock, no extra step. If the client has no domain yet, buy one in their business name during the handoff and add it to your invoice at cost.

know the bar

Where this wins and where it doesn't ๐ŸŽฏ

Gold for

Local service businesses, tradies, cafes and restaurants, salons and clinics, coaches and consultants, portfolio sites, and landing pages. Any brochure site whose job is to look credible and capture a lead, plus a simple appointment booker on top. That is most of what small businesses actually need.

Not the tool for

A complex web app with real logic: a marketplace, a booking-and-payments platform, a SaaS product, a big e-commerce catalogue with inventory. Anything needing a dedicated engineering team, or a $50k bespoke build. Scope those down to the marketing site they actually need first, or pass them on.

Selling this stack as something it isn't is the one sure way to burn a client and your reputation. Match the build to the real need and it holds up every time.


the fair question

Is the backend actually secure? ๐Ÿ”’

The honest knock on AI-built sites is that they look great on the front and hide a mess on the back. Here is the specific picture for Emergent, plus one caveat that applies to every AI builder.

Real production backendA React frontend, a FastAPI backend, and a managed database (MongoDB by default, or PostgreSQL via the Supabase integration). Reviewers who read the exported code call it clean, not spaghetti.
Proper login securityReal authentication (email, OTP, Google or GitHub) and, on the Supabase path, JWT sessions plus row-level security, so one user cannot read another's data.
Secrets stay secretClient keys are separated from server-only keys and stored as encrypted credentials in environment variables, not hardcoded into the frontend (the single most common vibe-coding leak).
Isolated containersEach app runs in its own isolated container, so one build cannot reach into another.
Certified platformEmergent is SOC 2 Type I and ISO 27001 certified, with SSO and audit logs. Nuance: that certifies the company, not automatically the specific app you generate.
You own and can audit itEvery line syncs to your own GitHub repo in standard languages, so you or any developer can read exactly what it built. Not a black box.
The honest caveat: no AI builder makes an app automatically bulletproof. Independent 2026 research found a meaningful share of AI-generated code ships with at least one security issue (Veracode put it around 45%). Emergent's managed backend and testing agents put it well ahead of raw prompt-and-pray coding, but before you collect real customer data: click through the acceptance tests, confirm login actually gates the private screens, and if anything sensitive is involved, have a developer skim the repo. That is just good practice with any build, AI or not.

The honest cost math ๐Ÿ’ฐ

Your side is about $20/mo of tools (Emergent Standard, 100 credits: 50 keep one site live, 50 to build and iterate) plus a free Claude plan for the strategy and copy. More client sites means more credits, so at real volume you move to a higher plan or buy top-ups (cheap per credit), but the per-site tool cost stays small.

Per client siteWhat businesses payWhat it costs you
A clean multi-page marketing site$2,000โ€“10,000 (Forbes Advisor)~$20/mo of tools, about an hour of your time
Domain$10โ€“20/year (billed to client)$0 (free DNS record)
Hosting and SSLbundled in your pricepart of the ~$20/mo credit plan
Realistic sell pricea few hundred to a few thousandthe gap is yours

The tools cost about $20/mo, the output is worth thousands, and a small monthly for hosting and edits makes the income compound. Just sell what this genuinely is, a clean small-business site, and it holds up every time.


Quick answers ๐Ÿ’ฌ

Do I need to code?No. Claude writes the plan and copy in plain chat, Emergent writes and hosts the code. Your job is describing the business well and clicking through the result like a customer would.
Do I need a paid Claude plan?No. The free plan runs the whole strategy and copy step. Pro (~$20/mo) just buys longer conversations if Claude is already in your daily workflow.
How many credits per site?A marketing site is light. Free (10 credits) builds a preview; Standard (100/mo) deploys one live (50/mo) with room to iterate. Specific prompts and free rollbacks are your best economy.
What if the first build looks wrong?Normal. Expect ~80% right, then 2 to 4 iteration prompts. If it keeps drifting, roll back to the last good checkpoint or restart with a tighter spec instead of patching forever.
Follow @cindiezhu for more Claude workflows that actually make you money and save you time ๐ŸŒฑ
Want more like this? Follow @cindiezhu for daily AI you can actually use.