# I launched Vibecode It Yourself

> I launched Vibecode It Yourself, a subscription audit that helps you replace paid apps with focused software built by your coding agent.

Author: [Flavio Copes](https://flaviocopes.com/about/) | Published: 2026-07-30 | Topics: [News](https://flaviocopes.com/tags/news/) | Canonical: https://flaviocopes.com/i-launched-vibecode-it-yourself/

Today [Pieter Levels posted](https://x.com/levelsio/status/2082477538966675867) that he had already seen free, open-source replacements for Wispr Flow, Granola, and WHOOP.

[Rob Hallam replied](https://x.com/robj3d3/status/2082478371825758688):

> someone should make a version that shows you subscriptions you pay for that you can vibecode yourself

I saw that reply and thought: yes. That should exist.

I built it in 10 minutes.

Five minutes later, after buying the domain name on Cloudflare, I deployed [Vibecode It Yourself](https://vibecodeityourself.com/) there.

It is a subscription audit, but optimistic.

Instead of only showing which subscriptions you forgot to cancel, it shows which ones you could replace with focused software of your own.

## What Vibecode It Yourself does

The site contains 33 personal and company subscriptions.

They include products like Linktree, Calendly, Typeform, Buffer, Notion, Trello, Slack, Loom, Mailchimp, Bitly, Plausible, Intercom, Statuspage, Zapier, Canny, HubSpot, Harvest, and Expensify.

You can search by app, category, or suggested stack. You can also filter the catalog by the estimated effort:

- **Tonight** for the smallest replacements
- **Weekend** for projects with a little more depth
- **Side quest** for the ambitious ones

Each card shows an estimated monthly price, a rough build estimate, the narrow product idea, and a possible technical stack.

Click **I could build this** and the site adds it to your audit.

The receipt at the top updates with your potential yearly savings, the number of apps selected, and the estimated build time.

There is also a **Select the no-brainers** button. It picks the ten smallest projects in one click.

No bank login is needed. There is no account and nothing to import.

You are not giving a financial company access to your transactions. You are looking at a list and recognizing the subscriptions you already know you have.

## You are not cloning entire companies

This is the most important part of the idea.

You should not ask a coding agent to "build Notion" or "build Slack."

Those are giant products built by large teams over many years. Their difficult parts are not the buttons you can see in a screenshot.

The useful question is smaller:

> What part of this product do I use every week?

Maybe you pay for Notion, but all you need is a private team wiki with nested Markdown pages, search, page owners, and a small revision history.

Maybe you pay for Calendly, but you are one consultant offering 30-minute and 60-minute calls. You need availability rules, a booking page, calendar sync, and confirmation emails. You do not need team routing, payments, or five calendar providers.

Maybe you pay for Canny, but you only need one feedback board. Customers post requests, other customers vote, and you move items through planned, in progress, and complete.

Maybe you pay for Statuspage, but you only need public service health, incident updates, scheduled maintenance, and an email list.

Maybe you pay for Linktree and need a profile, a short bio, and an ordered list of links.

That last one is a very different project from rebuilding a general link-in-bio platform for millions of people.

This is where vibecoding becomes interesting.

A general SaaS must work for many industries, team sizes, countries, billing plans, and workflows. Your version only needs to work for you.

You can remove most of the product before you write the first line of code.

## Every replacement has a detailed build prompt

The app idea is only the start.

Every card has an **Open full build prompt** button. It gives you a detailed specification ready to paste into a coding agent.

I did not want these prompts to say "build a clone of X."

That usually creates a nice landing page, a generic dashboard, and a few buttons that do not lead anywhere.

An agent can build the wrong product very quickly.

The prompts define:

- the exact person the app is for
- the outcome that person wants
- the main user journeys
- the screens and their actions
- the small set of features that must work
- the data model and state transitions
- authentication and permission boundaries
- external integrations
- security and privacy rules
- acceptance scenarios and tests
- deployment and recovery requirements
- explicit non-goals

The non-goals matter as much as the features.

For example, the focused Linktree replacement does not include multi-tenant billing, a theme marketplace, or enterprise analytics.

The focused Trello replacement does not include workspace billing, Power-Ups, portfolio views, or a complex automation system.

The focused Plausible replacement does not include attribution modeling, ad integrations, or visitor profiles.

Scope is what makes these projects buildable.

The prompt also asks the agent to plan before writing code. It must identify pages, endpoints, database tables, permissions, background jobs, and external services.

Then it must build a complete vertical slice with persistent data, validation, error handling, and tests.

Not a mockup. Not a pile of disconnected components.

A working first version.

## There are 33 starting points

The launch catalog covers six areas:

- personal tools
- productivity
- marketing
- sales
- support
- operations

Some projects are tiny by design.

A personal link page, image compressor, short-link service, form, or file-transfer page can be a great first project.

Others need more care.

A calendar booking system must handle time zones and double bookings. A newsletter tool must handle consent, unsubscribes, bounces, and sending reputation. A support inbox must keep private notes separate from customer replies.

The larger ideas are marked as side quests because the visible interface is only part of the work.

Slack, Intercom, Airtable, HubSpot, and Zapier are not casual evening projects, even when you cut their scope aggressively.

The effort labels are rough estimates, not promises. Your experience, requirements, coding agent, and review process will change the real time.

The point is to compare scope, not to start a stopwatch.

## Not every subscription should be replaced

I am not saying SaaS is bad.

I love paying for software that solves a hard problem well.

When a product handles money, sensitive data, critical communication, or a complicated integration, the subscription can be the cheapest part of the system.

You are also paying for maintenance, backups, security updates, support, uptime, and somebody else waking up when production breaks.

If you build your own version, those jobs become yours.

This is why the site says:

> Keep the hard stuff. Build the tiny stuff.

Do not replace a reliable business system only because its interface looks easy to copy.

Replace the products where your real workflow is much smaller than the product you are renting.

A form with seven questions does not need to become a form-building company.

A private directory does not need enterprise permissions and a marketplace.

A status page for three services does not need to support thousands of customers.

That is the sweet spot.

## Why this is possible now

For a long time, subscribing was the obvious choice.

Building a small internal tool meant design, frontend code, backend code, authentication, hosting, testing, and maintenance. Even a narrow product could take weeks.

Paying $10 or $20 per month was an easy decision.

Coding agents changed the calculation.

They do not remove the need for product decisions or review. They do make implementation much faster, especially when the requirements are clear and the scope is small.

You can now describe a workflow, inspect the plan, let the agent implement it, and spend your time reviewing what matters.

This makes a new kind of software practical: software built for one person or one small team.

It does not need a pricing page. It does not need onboarding for strangers. It does not need to become a startup.

It only needs to solve your problem.

The result can fit your workflow better than a general SaaS because it has no reason to serve everyone else.

## How I built Vibecode It Yourself

I built the first version in 10 minutes.

The site is intentionally small:

- Astro generates one static page
- TypeScript contains the subscription catalog and build specifications
- vanilla JavaScript handles search, filters, selections, the savings receipt, prompt dialogs, and dark mode
- plain CSS handles the design
- Cloudflare Pages hosts the result

There is no application server, database, account system, or bank integration.

The entire audit runs in your browser. Selecting a subscription only updates the page in front of you.

The catalog is data-driven. Each subscription starts with a focused definition like this:

```ts
{
  name: 'Linktree',
  projectName: 'Linkhouse',
  effort: 'easy',
  time: '30 min',
  audience: 'one creator managing a single public profile',
  goal: 'publish a fast, branded bio page and understand which links get clicked',
  stack: ['Astro', 'Alpine', 'D1'],
}
```

The real entry also contains features, data entities, integrations, security guardrails, and non-goals.

A second set of data adds the app-specific journeys, screens, business rules, and acceptance scenarios.

One shared function turns all of that into the full prompt.

This means I can add another subscription without designing another page or inventing a new prompt format.

The interface, calculator, and prompt generator all read from the same source.

## Start with one no-brainer

Do not select all 33 apps and spend the next six months rebuilding your life.

Pick one small subscription that annoys you.

Choose something where you use one narrow feature and ignore the rest of the product.

Open the prompt. Read it. Delete anything you do not need. Add the small details that make the tool yours.

Then give it to your coding agent and build the first version.

If it works, cancel one subscription.

If it does not, you still learned more than you would by watching another tutorial.

Keep the hard stuff. Build the tiny stuff.

[See what you can replace](https://vibecodeityourself.com/).