The Free Guide: A Small Paid App, Built With Claude, Driven From Your Phone (Built 2026-09-20)

Every step below comes from a file that was checked. The file is named on the step. Nothing here is from memory. Where a step could not be confirmed from those files, the line says VERIFY LIVE BEFORE FILMING instead of guessing.

---

What you get (read this first, it is honest)

This is the free way to put one small paid app in front of real buyers: one screen, one job, one price, with a checkout inside it, an icon on a phone home screen, and a one page site in front of it.

The one paid door, named with its price

There is exactly one thing in this whole guide you have to pay for, and it is the Claude plan itself.

Everything else in this guide is free for real. Say that plainly to yourself: the free route takes your time instead of your money. You will write the app with one message and then fix the last fifth of it yourself, one change at a time.

---

Step 1: the phone really does drive it

Source for this whole step: updated-tools-ideas-3-and-4-2026-09-20.md, row a, quoting code.claude.com/docs/en/mobile, checked 2026-09-20, and card-video4-app-from-phone-2026-09-20.md, step 1.

Here is the part every "I built it on my phone" video leaves out, said straight: the code does not run on the phone. The phone is a remote control. That is Anthropic's own framing on their own page, and it is better than the fake version, because the honest one actually works.

Three real doors, in their words:

And the three things the phone cannot do, straight off the same page, so nobody wastes an hour finding out:

• /resume does not work from the phone.
• /plugin does not work from the phone.
• You cannot pick bypass mode from the app.

That is the whole honest picture. The phone genuinely drives it. The machine still does the work.

Is it safe to let the AI onto your machine and into your files, driven from a phone?

Yes, when you set the fences first. The AI reads a plain text rules file sitting in the folder you work in, and it follows it every single time.

Put it in the folder the app lives in and write the fences in plain words. For example:

• Never touch, move or delete anything outside this folder.
• Never delete a file. Move it to a folder called Old instead.
• Never open my keys, my passwords, or any settings file that holds them.
• Never send anything out: no email, no upload, no message, without asking me first.
• Never publish anything anywhere without showing me first.
• Always work on a copy, never the original.

That file is the one to have open on screen. It is the file that says what the AI may never touch. (Source: card-video4-app-from-phone-2026-09-20.md, the safety line inside step 1.)

---

Step 2: build the app with one message

Source for this whole step: updated-tools-ideas-3-and-4-2026-09-20.md, rows a and d, and card-video4-app-from-phone-2026-09-20.md, step 2.

The shape is the whole trick: one screen, one job, one price. Not a product with a menu. One screen that does one thing somebody will pay for.

The site in front of it is shaped the same way. It is the hallway with one door: one page, one button, no menu, no about page. (Source: updated-tools-ideas-3-and-4-2026-09-20.md, row d, from Nathan Nazareth's shape as described there.)

Expect the first output to be about 80 percent right. The last 20 percent is yours, and you get it with one change per prompt, never a list of five changes in one message. (Source: card-video4-app-from-phone-2026-09-20.md, step 2.)

The build message, written out in full

Paste this in and swap the things in square brackets.

Build me one small paid web app and its landing page, both in this folder.

THE APP
One screen. One job. One price.
The one job it does is: [WHAT IT DOES, IN ONE SENTENCE, NO EXTRAS]
The person it is for is: [WHO THEY ARE AND WHAT THEY ARE STUCK ON]
The price is: [YOUR PRICE, SOMEWHERE BETWEEN $27 AND $97]

Rules for the app:
- One screen only. No menu, no tabs, no settings page, no login, no account.
- Plain HTML, CSS and JavaScript in one folder. No build step, no framework,
  no package to install.
- Phone width first. It has to look right on a phone before it looks right on
  a laptop, because that is where it gets used.
- Whatever the person types stays in their own browser. Nothing is uploaded
  anywhere. No tracking, no analytics, no outside calls.
- Put a single button where the checkout link will go, with the text
  CHECKOUT LINK GOES HERE, so I can paste my payment link in one place.
- Do not put any key, token or password anywhere in the app.

THE LANDING PAGE
One page. One door. Build it as a hallway with one door: the visitor arrives,
there is exactly one thing to do, and that is the buy button.

Rules for the page:
- No menu. No about page. No footer links going anywhere else.
- Top of the page: the one sentence that says what it does and who it is for.
- Then three lines on what the person gets. Plain words, no marketing language.
- Then the price, written plainly, and the one button.
- Then nothing else.
- Same one checkout placeholder: CHECKOUT LINK GOES HERE.

Never do any of these:
- Never invent a testimonial, a review, a user count or an earnings figure.
- Never add a newsletter box, a chat widget, a cookie banner or a tracker.
- Never touch, move or delete any file outside this folder.
- Never write a key, a token or a password into any file.

When it is done, print the list of files you wrote and tell me the one command
that opens it in a browser.

The change prompts, one at a time

After the first version, you fix the last 20 percent with single changes. One per message. Examples of the right size:

Make the button full width on a phone screen and leave everything else alone.

Change the headline to: [YOUR NEW HEADLINE]. Change nothing else.

The result box is too small on a phone. Make it taller. Do not touch anything else.

Move the price above the button. That is the only change.

The reason for one change per prompt: when you send five, you cannot tell which one broke the screen. When you send one, you always can.

---

Step 3: the checkout, and the comparison nobody does

Source for this whole step: updated-tools-ideas-3-and-4-2026-09-20.md, row b, quoting stripe.com/pricing, lemonsqueezy.com/pricing, polar.sh and gumroad.com/pricing, all checked 2026-09-20, and card-video4-app-from-phone-2026-09-20.md, step 3.

No code at all. A Stripe Payment Link is made in the Stripe dashboard and pasted into your app like any normal link.

The steps, in plain words:

1. Open your Stripe dashboard and make a new Payment Link.
2. Name the product what the app is called, and write the one line of what the buyer gets.
3. Set the price. This guide films it at $47, inside the $27 to $97 band.
4. Set it as a one time payment, not a subscription, unless you actually meant a subscription.
5. Set where the buyer lands after paying, so they get the thing they just bought.
6. Copy the link.
7. Paste it into the app and the landing page, in the one spot the build message left as CHECKOUT LINK GOES HERE.

VERIFY LIVE BEFORE FILMING: the exact button names and the order of the screens inside today's Stripe dashboard are not written in any checked file. Make the link on camera and read the buttons off the screen.

The same $47 sale, four ways

(All four rows: updated-tools-ideas-3-and-4-2026-09-20.md, row b, from each tool's own pricing page, checked 2026-09-20.)

The trade off, said plainly. Stripe is the cheapest and you keep the most, but the sales tax is yours to deal with. Lemon Squeezy and Polar take more per sale and handle the tax for you. That is the whole choice, and nobody in the source videos says it out loud.

---

Step 4: get it onto a phone, with no app store

Source for this whole step: updated-tools-ideas-3-and-4-2026-09-20.md, row c, quoting appilix.com/pricing, checked 2026-09-20, and card-video4-app-from-phone-2026-09-20.md, step 4.

Here is the fact worth sending to a friend: an icon on somebody's phone home screen costs nothing, and there is a tool charging $119 for the same thing.

Door one: Add To Home Screen. Free, needs nothing.

1. Open your live app in the phone's browser.
2. Open the browser's share or menu button.
3. Tap Add To Home Screen.
4. The icon is now on the home screen and it opens full screen like an app.

VERIFY LIVE BEFORE FILMING: the exact wording and menu position of Add To Home Screen on today's iPhone and Android browsers are not written in any checked file. Tap it on camera and read the screen.

Door two: PWABuilder. Microsoft's, free.

It wraps a web app into an Android package. Free.

VERIFY LIVE BEFORE FILMING: PWABuilder's own docs page did not load on 2026-09-20, so it is marked as free but not price checked, and the click by click steps are not claimed here (source: updated-tools-ideas-3-and-4-2026-09-20.md, "What is missing").

The correction: Appilix is not free

Its own pricing page says so, checked 2026-09-20:

You do not need an app store and you do not need $119 to have an icon on someone's phone.

---

Step 5: the first buyers, and what happens after you ship

Source for this whole step: updated-tools-ideas-3-and-4-2026-09-20.md, rows e and f, and card-video4-app-from-phone-2026-09-20.md, step 5.

Finding the first buyers is not posting

Every 2026 playbook in the checked files says the same thing, and it is not posting into a feed. It is this:

1. Go where somebody already wrote the complaint. The exact problem your one screen fixes, written out by a real person who was annoyed enough to type it.
2. Answer it properly. A real answer, the one you would give a friend, whether or not they ever buy.
3. Mention the app once. Once. At the end. As the thing you built because of that exact problem.

(Sources, all with their own view counts as checked: Starter Story, "How This $250K/Month SaaS Got Its First 100 Users", youtu.be/l4WEqPX52Cg, 114,949 views, uploaded 2026-03-11 · Adam Robinson, "If I Started a SaaS in 2026, I'd Do This (0 to 1,000+ Users)", youtu.be/r47Q4Cfit5s, 84,889 views, uploaded 2025-12-31 · Rob Walling, "How to Find Your FIRST 10 SaaS Customers", youtu.be/Ki3qBvIRo6A, 53,920 views.)

The proof wall you can click

TrustMRR pulls its revenue numbers straight from Stripe, RevenueCat, Superwall and Creem, and refreshes hourly (trustmrr.com, checked 2026-09-20). That means you can look at a real app's real number instead of trusting somebody's screenshot.

The three lines the source videos skip

This is the question under the source video that nobody answers, and here is the answer in three lines:

1. The keys never go in the app. They live in the host's environment settings. Not in a file, not in the code, not in the page.
2. The payment page is Stripe's, not yours. You never hold a card number, because the card is typed on Stripe's page.
3. One hour a week on the inbox is the whole maintenance job at this size. That is it.

(Source: updated-tools-ideas-3-and-4-2026-09-20.md, row f.)

After you ship: get a real guard, not a promise

Two things, and both are free.

One: hand your AI the protections list. Take a written list of the protections a real live app has (the one used in the video is written from a working app's own repo, with no secrets in it) and give it to your own AI with one instruction:

Here is a list of the protections a real live app has in place.

[PASTE THE PROTECTIONS LIST]

Go through my app and implement the same things on mine. For each item on the
list, tell me one of three things:
1. Already done, and where it is done.
2. Not done, and here is the change (show me the change before you make it).
3. Does not apply to an app this size, and why.

Rules:
- Never put a key, a token or a password into any file. Keys live in the host's
  environment settings.
- Show me every change before you make it.
- Do not invent a protection that is not on the list.
- If something on the list would break the app, say so instead of doing it.

Two: run a security review on your own app. Run it with a model that will actually review code and say what it finds. The video runs it twice on the demo app, once with Kimi and once with DeepSeek Flash 4.1 uncensored, and fixes what comes back in the same pass. (Source: card-video4-app-from-phone-2026-09-20.md, step 5 after-you-ship block.)

The security review prompt, written out in full:

You are reviewing a small paid web app for security holes. Be blunt. I would
rather hear a real problem than a polite summary.

Here is the app: [PASTE THE FILES, OR POINT AT THE FOLDER]

Go through it and find:
1. Any key, token, password or secret written into a file, anywhere.
2. Anything the app sends out: what it sends, where it sends it, and whether
   the person using it would expect that.
3. Any place someone's typed input is put straight into the page, the URL, or a
   request without being cleaned first.
4. Anything a visitor could change on their side to get the paid thing for free.
5. Anything that stores personal information, and whether it needs to.
6. Any outside script or library the page loads, and what happens if that
   outside site is hacked.

For every problem you find, give me:
- What it is, in plain words.
- How bad it actually is for an app this size, honestly. Do not inflate it.
- The exact change that fixes it.

Then tell me the ONE thing to fix first.

Rules:
- Do not invent a vulnerability to have something to say. If it is clean, say
  it is clean.
- Do not tell me to add a big framework or a paid service.
- Plain words, no jargon.

---

The money maths (real numbers only)

Source: card-video4-app-from-phone-2026-09-20.md, money math, and updated-tools-ideas-3-and-4-2026-09-20.md, row b.

At $47 through Stripe you keep $45.34 per sale.

Thirty sales a month is what Erik Cupsa's $1,400 a month looks like in real life. He said that figure on his own video in 2026: $1,400 a month recurring, $16,500 lifetime, built in 23 days (youtu.be/KKWS0bSiCT4, uploaded 2026-06-15, 395,264 views).

The ceiling shown, each with its year and whose money it is:

Those are their numbers, in their years, said on their own videos. They are not a promise about yours.

The price band: $27 to $97. The reason the band is low and not high is a comment with 11 likes under one of the source videos, saying nobody is paying a huge invoice to somebody they do not trust yet. (Source: card-video4-app-from-phone-2026-09-20.md, "The thing built before he scripts", price row.)

---

Links

TRACKED LINK NEEDED BEFORE PUBLISHING on every link that leaves with the video: the free guide link, the community link, the sponsor link and the ending card link each go out as a named tracked address on Reis's own domain, added to the tracker first, never the raw destination. (Source: card-video4-app-from-phone-2026-09-20.md, "Links (metric 11)".)

The video links and the pricing pages quoted in this guide are citations shown on screen, not published links.

---

The free community tool

The App Starter is free inside the free community. Join to get it.

It is the same method as this guide with the thinking already done: you type the one job of your app, who it is for, the price and your name, and it hands you back the exact build message, the change prompts, the checkout checklist, the install steps, the after you ship checklist and the money maths for your own price. (Source: card-video4-app-from-phone-2026-09-20.md, approved line 2026-09-20 late: the app starter is the free community tool, and community tools are free inside the free community.)

---

Open items before this guide is published