GuidesArtifacts

The Claude mini app starter kit

Eleven working apps, built by describing them.

Calculators, trackers, dashboards, quizzes and tools you can build in Claude without writing code. Eleven copy-paste prompts, the prompt that improves any of them, and how to turn your own repeated tasks into apps.

Free. No email required, nothing to download.

What you get
  • Eleven copy-paste prompts, each producing a working interactive app
  • The one setting you have to switch on first, and which plans support what
  • The review prompt that turns a demo into something usable
  • Four levels to take any app through, in the order that actually works
  • The prompt that finds your own app ideas instead of copying these

Before you start

You do not need JavaScript, React, or any idea how to deploy a website. You describe what you want and Claude builds an interactive artifact that runs next to the conversation.

One setting first, because nothing below works without it. Turn on Cloud code execution and file creation in Settings, then Capabilities. On a Team or Enterprise plan an owner switches it on in the organisation settings.

  • All eleven work on the free plan. Artifacts are available on Free, Pro, Max, Team and Enterprise.
  • Saving data between sessions needs Pro or above. Which matters for the trackers and anything you want to come back to. On free, treat them as calculators rather than records.
  • Connecting other apps needs Pro or above. There is a separate toggle for artifact connectors in the same Capabilities settings.
  • Sharing one costs you nothing. No API keys, and when someone else uses your app the usage counts against their plan, not yours.

Then the loop: pick an app, copy the prompt, paste it in, use the thing it builds, and tell it what to change. You do not need the prompt to be perfect, because the first version is a starting point you talk to.

Make the interface simpler.
Add another input for X.
Change the calculation to work like this.
Add a chart showing the results.
Make this easier to use on mobile.
Let users save previous results.

That is the part people do not expect. You are not writing a specification, you are correcting something that already runs. Most of these get genuinely useful on the third or fourth thing you ask for.

ROI calculator

A calculator that tells you whether something is actually worth it, instead of you squinting at a spreadsheet.

The fastest one to build, and the one that makes the point in about ten seconds.

Useful for:

  • Marketing campaigns and paid ads
  • Software and tool purchases
  • Hiring, or contractor versus employee
  • Automation projects
  • Equipment
  • Sponsorships
01

Build the ROI calculator

Paste it as-is. You are not writing a spec, you are starting a conversation you can correct.

The prompt
Build me an interactive ROI Calculator as a Claude Artifact.

Do not just explain how to calculate ROI. Create the working app.

The user should be able to enter:
- Initial investment or cost
- Ongoing monthly cost
- Revenue generated
- Monthly savings
- Time period in months

Automatically calculate:
- Total cost
- Total financial benefit
- Net return
- ROI percentage
- Payback period where possible

Display the results in a simple visual dashboard.

Clearly separate revenue generated from cost savings so they are not accidentally
double counted.

Include a short plain English summary, for example: "Over 12 months this
investment costs X and generates or saves Y, an estimated return of Z percent."

Add input validation, sensible empty states and a reset button.

Make the design clean and easy for a non-technical person to use. Do not leave
sample data in place once the user starts entering their own numbers.
Result

A working calculator you can hand to someone else, built from one paragraph.

Level it up. Ask for three scenarios: conservative, expected and optimistic. One number invites you to believe it. Three make you ask what happens if you are wrong, which is the actual question.

Pricing calculator

Works out what to charge for a project, from your own rates rather than a number someone invented on a blog.

02

Build the pricing calculator

Everything is calculated from the rates you put in, which is what makes the quote defensible.

The prompt
Build me an interactive Pricing Calculator as a Claude Artifact.

It should help a freelancer or consultant calculate a project price from their own
numbers.

Ask the user for:
- Desired hourly or day rate
- Estimated hours or days
- Additional expenses
- Revisions included, and cost per extra revision
- Rush delivery percentage
- Profit margin or buffer
- Optional tax percentage

Calculate base project cost, additional costs, rush fee if selected, buffer,
subtotal, tax where applicable, and the final quote.

Show a transparent breakdown so the user can see exactly how the price was built.

Use toggles for optional fees rather than forcing every input.

Also suggest three packages based on the calculated base price: basic, standard
and premium.

Do not claim any price is an objectively correct market rate. Everything should be
calculated from the user's own inputs.

Make it clean and usable on a phone.
Result

A quote with a visible breakdown, and three package options you did not have to build by hand.

Level it up. For creators, add paid usage rights, exclusivity, cross-posting, extra deliverables and raw footage. For freelancers, add scope complexity and project management time. Those are the fields people forget and then absorb for free.

Resume checker

Paste a CV and a job description and get a structured comparison instead of a vague feeling about whether to apply.

03

Build the resume checker

The instructions against inventing experience matter more here than anywhere else in this guide.

The prompt
Build me an interactive Resume Checker as a Claude Artifact.

Give it two input areas: the resume or CV, and the job description.

When both are provided, analyse the match and show:
- Key requirements in the job description
- Which requirements appear supported by the resume
- Important requirements not clearly demonstrated
- Relevant experience that is currently buried or badly explained
- Keywords and terminology worth considering
- Weak or vague bullet points
- Suggested improvements

Organise it into: strong matches, potential gaps, resume improvements, and
questions to ask myself before editing.

Do not invent qualifications, experience, achievements or metrics, and never
recommend adding something the candidate has not actually done.

Do not produce a fake ATS probability score. Use a transparent checklist showing
which requirements are evidenced and which are not.

When you suggest a rewritten bullet point, show the original next to the suggested
version and explain what improved.

Keep the interface simple and easy to scan.
Result

An honest checklist of what the CV evidences, and rewrites shown next to the originals so you learn the pattern.

Level it up. Add a seniority selector: junior, mid, senior, manager, executive. The same CV against the same job gets different feedback depending on the level, because what counts as evidence changes.

Lead scoring tool

Decides which prospects get your attention first, using criteria you set rather than a black box.

04

Build the lead scoring tool

The weighting is the real output. Arguing with yourself about the weights is where the thinking happens.

The prompt
Build me an interactive Lead Scoring Tool as a Claude Artifact.

Let me define my own qualification criteria and weighting. Start with editable
criteria such as company size, budget, industry fit, decision making authority,
urgency, problem fit, engagement level and existing relationship.

Let me assign each criterion a weighting and a score.

Calculate an overall weighted score out of 100 and categorise leads as high,
medium or low priority.

Show exactly how each score was calculated.

Allow multiple leads so they can be compared in a table, let me sort by score, and
include a notes field.

Do not present the score as a prediction of whether someone will buy. It is a
prioritisation system based on criteria I chose.

Make it clean and suitable for someone working in sales, consulting or an agency.
Result

A ranked, sortable list of prospects with the reasoning visible, not a mystery number.

Level it up. Describe your actual ideal customer, then ask it to replace the generic criteria with qualification criteria built around that profile. Generic criteria score generic leads.

Web page audit tool

A structured review of one page, which is the honest version of this app.

Claude is not crawling your site from a chat, so an app that pretends to is a worse app. Give it the page and it can do real work on it.

05

Build the page audit tool

One page at a time, with its goal and audience supplied. That is what makes the findings specific rather than a generic SEO checklist.

The prompt
Build me an interactive Web Page Audit Tool as a Claude Artifact.

The user pastes in the page copy, the page title, the meta description if they
have it, the primary goal of the page, the target audience, and the target keyword
if relevant.

Audit across: clarity, value proposition, search intent, content, conversion,
trust and structure.

For each area, show the findings, explain why they matter, and suggest specific
improvements.

Finish with a prioritised action list: fix first, fix next, nice to improve.

Do not invent technical SEO issues that cannot be determined from what was
provided. Where something needs technical data you do not have, say so instead of
guessing.
Result

A prioritised list of what to change on that page, and an explicit note on what cannot be judged from copy alone.

Level it up. Feed it more evidence and it stops being an opinion: a screenshot, a Search Console export, GA4 landing page data, competitor pages. The same prompt becomes a landing page CRO checker, a product page reviewer or a homepage messaging reviewer depending on what you tell it the page is for.

Project prioritisation tool

The cure for a list of twenty-seven ideas where everything is apparently urgent.

Good enough to use live in a planning meeting, which is the real test.

06

Build the prioritisation tool

Build it before the meeting, then change the weightings in front of everyone. Watching the ranking move is a faster argument than a slide.

The prompt
Build me an interactive Project Prioritisation Tool as a Claude Artifact.

Let the user add multiple projects. For each one collect name, expected impact,
effort required, urgency, strategic importance, confidence, estimated cost and
dependencies.

Let the user adjust the weighting of each factor, and calculate a transparent
prioritisation score.

Display the projects as both a ranked table and an impact versus effort matrix.

Categorise them as do now, plan next, investigate, or low priority.

Do not present the score as objective truth. Make it clear the ranking reflects
the user's inputs and chosen weightings.

Let projects be added, edited and removed.

Make it visual and easy to use during a planning meeting.
Result

A ranked list and a matrix, both of which update when you change what you say matters.

Level it up. Give it templates for different jobs: marketing weights revenue potential, reach, effort and confidence; product weights customer impact, strategic fit, engineering effort and risk; personal weights importance, urgency, effort and enjoyment. One app, three uses.

Client onboarding app

Replaces the seventeen onboarding emails that all begin with sorry, one more thing.

07

Build the onboarding app

Then send it to the next client instead of the email chain.

The prompt
Build me an interactive Client Onboarding App as a Claude Artifact.

It should collect everything needed before starting a new client project, in
sections:

Business: company name, website, contact details.
Project: service required, main objective, expected outcome, deadline, budget if
relevant.
Audience: target customer, their important problems.
Brand: brand voice, examples and references, things to avoid.
Access and assets: tools or accounts needed, documents needed, brand assets.
Communication: main contact, preferred method, meeting cadence.

Show onboarding progress at the top, and highlight unanswered required questions
before submission.

At the end, produce a structured onboarding summary that can be copied out.

Make it feel welcoming and simple rather than like a large corporate form.
Result

One link that collects everything, and a summary you can paste straight into your project notes.

Level it up. Tell it what you actually sell and the questions change completely. "I run an SEO consultancy, adapt this for SEO clients" produces a different form than "I make short-form UGC for brands". Generic onboarding forms are why clients skip questions.

The custom version

Guides like this are the generic version.

Everyone reading this page gets the same system. We build content systems that learn your voice from what you have actually published, know who you are writing for, and make every piece better because of how the last one performed. Then we either hand it to your team or run it for you.

Our systems drive results
267K+
Followers built
8M
Views in 7 months
4
Channels in parallel

Interactive quiz

Turns anything you are teaching or learning into a quiz that gives feedback as you go.

Especially good for:

  • Courses and staff training
  • Exam revision
  • Lead magnets
  • Employee onboarding
  • Product education
08

Build the quiz app

Replace the topic, and decide whether the questions come from your material or from Claude's own knowledge. That choice matters.

The prompt
Build me an interactive Quiz App as a Claude Artifact.

Create a quiz on this topic: [INSERT TOPIC]

I want to be able to provide the questions myself, or have you generate them from
material I give you.

Include one question at a time, multiple choice answers, immediate feedback, an
explanation after each answer, a progress indicator, a final score and a results
summary.

At the end, show which topics were answered correctly, which ones the user
struggled with, and suggested areas to revise.

Include a restart button and randomise question order where practical.

Keep the interface playful but clean.

Do not write factual questions from information I have not provided unless I
explicitly ask you to use your own knowledge.
Result

A working quiz with explanations, and a summary of what the person actually needs to revise.

Level it up. Ask for adaptive difficulty: harder questions after a run of correct answers, simpler ones and more explanation when someone is struggling. Then have it turn the results into a short revision plan. A score is a dead end, a plan is not.

Budget planner

A budget you can actually see, with a what-if mode that a spreadsheet screenshot will never give you.

09

Build the budget planner

The scenario mode is the reason to build this rather than use a template.

The prompt
Build me an interactive Monthly Budget Planner as a Claude Artifact.

Let the user enter income (main and additional), fixed expenses (housing,
utilities, insurance, subscriptions, loan payments), variable expenses (groceries,
transport, eating out, entertainment, shopping, other) and goals (monthly savings
target, optional financial goal).

Automatically show total income, total expenses, remaining amount, savings amount,
percentage of income spent and percentage saved.

Include a visual breakdown by category and let users add custom categories.

Add a scenario mode where changing one expense instantly shows the impact on the
month.

Keep every calculation transparent.

Do not give personalised financial advice, and do not claim any particular
allocation is right for everyone.
Result

A clear picture of the month, and instant answers to what if rent goes up by 200.

Level it up. The what-if mode is the whole build. What if rent rises, what if these subscriptions go, what if there is another 500 a month coming in. Those are the questions people actually have, and a static budget cannot answer any of them.

Habit tracker

A light tracker for whatever you are trying to do consistently, without the guilt design.

Versions worth making:

  • Running or training tracker
  • Study tracker
  • Content posting tracker
  • Reading tracker
  • Sales outreach tracker
10

Build the habit tracker

Note the instruction about language. Most trackers are built to make you feel bad, which is why people abandon them.

The prompt
Build me an interactive Habit Tracker as a Claude Artifact.

Let users add habits, choose which days each habit should happen, mark them
complete, edit or delete them, see today's habits and see weekly progress.

Show completion percentage, current streak, longest streak and weekly consistency,
plus a simple calendar-style history.

Avoid guilt-heavy messaging when someone misses a habit. Use neutral language, for
example "3 of 5 planned sessions completed this week."

Include an optional notes field for each day.

Keep the interface minimal and satisfying to use.
Result

A tracker that reports what happened without editorialising about it.

Level it up. Add a weekly reflection that asks three questions every Sunday: what worked, what got in the way, what changes next week. On Pro and above the artifact can keep your entries between sessions, which is what makes a tracker worth using past week one.

Content idea app

A structured idea generator where you control the inputs, rather than a button that says generate ideas.

11

Build the idea app

The controls are the point. Ideas without a platform, a goal and a format attached are not ideas yet.

The prompt
Build me an interactive Content Idea Generator as a Claude Artifact.

Let the user choose:
- Platform: TikTok, Instagram, YouTube, LinkedIn, X
- Goal: reach, education, authority, engagement, conversion
- Format: tutorial, list, story, opinion, before and after, mistake, comparison,
  case study
- Topic: free text
- Audience: free text

Then generate content concepts, each with an idea, a working hook, the core
takeaway, a suggested format and a CTA where one fits.

Add controls for safe versus experimental ideas, and for beginner, intermediate or
advanced audiences.

Include a button that regenerates only the hook without replacing the whole idea.

Include a favourites function so good ideas can be kept.

Do not produce ten minor variations of the same concept. Prioritise genuinely
different angles.
Result

Ideas with a hook, a format and a goal attached, and a favourites list that survives the session.

Level it up. Then give it your content pillars, brand voice, audience profile and the topics you have already covered, and ask it to build that context into the app. That is the difference between your idea generator and a generic one with a nicer button.

The prompt that improves any of them

Every app above comes out of the first attempt looking like a demo. This is how you get it to something a person would actually use, and it works on all eleven.

12

Review and fix your own app

Paste this straight after Claude builds the first version, before you start requesting features.

The prompt
Review the app you have just created as if you were a first-time user.

Look for:
- Unclear instructions
- Unnecessary inputs
- Missing functionality
- Broken or confusing interactions
- Weak mobile usability
- Poor empty states
- Places where the user would not know what to do next
- Calculations or assumptions that are not transparent

Give me the five highest impact improvements, then implement them.
Result

The gap between "look, Claude made an app" and something you would send to someone else.

Four levels, in this order

  1. 01
    Make it work

    Inputs, a calculation or action, a useful output. That is a finished app. Most people never need to go further.

  2. 02
    Make it yours

    Your criteria, your industry, your audience, your terminology, your workflow. This is where a generic tool becomes your tool.

  3. 03
    Give it memory

    For trackers and anything you return to, have it keep your entries between sessions. Available on Pro, Max, Team and Enterprise.

  4. 04
    Connect it

    An artifact can talk to apps you have connected to Claude, on Pro and above, with each person authorising their own connections. A prioritiser that writes to your task manager, onboarding that creates the project.

Do not start at level four. A connected version of an app nobody uses is just a more expensive way of not using it. Prove the simple version earns its place first.

Which one to build first

  • Normal office job. Project prioritisation tool.
  • Job hunting. Resume checker.
  • Freelancing. Pricing calculator.
  • Running a business. Lead scoring tool.
  • Making content. Content idea app.
  • Learning or teaching something. Interactive quiz.
  • Just want to see whether this is real. ROI calculator. It takes about ten seconds to understand, which makes it the right first one.

And you do not have to stop at eleven. Every one of these is the same shape: inputs, logic, a useful output, an interface. Once you can see that shape, most repeated tasks you do in a spreadsheet are candidates.

13

Find your own app

The most useful prompt here, because it stops you copying our list and starts you building for your own work.

The prompt
I do [TASK] repeatedly.

Help me decide whether this could be turned into a small interactive app.

Ask me questions about the inputs, the decisions, the rules and the outputs involved.

Then suggest three possible app designs, starting with the simplest useful version.
Result

Three designs for something only you needed, and an honest answer when a task is not worth an app.

Common questions

  • Do I need to know how to code? No. You describe what the app should do and Claude writes it. The skill you actually need is describing what you want and then correcting what you get, which is closer to briefing a contractor than programming.
  • Why can I not create artifacts? Almost always one setting. Cloud code execution and file creation has to be on in Settings, then Capabilities. On Team or Enterprise plans an organisation owner has to enable it first, which is why it sometimes cannot be switched on from your own account.
  • Do these work on the free plan? Yes, all eleven. Artifacts are available on Free, Pro, Max, Team and Enterprise. What free does not include is keeping data between sessions or connecting other apps, both of which need Pro or above, so on free the trackers work as calculators rather than records.
  • Can my app remember what I entered last time? On Pro, Max, Team and Enterprise, yes. Artifacts can retain data between sessions on those plans, which is what makes the habit tracker and the budget planner worth using past the first week. On free, the app resets.
  • Can I share these with other people? Yes, and it costs you nothing. There are no API keys involved, and when someone else uses your app the usage counts against their own plan rather than yours. That makes the quiz and onboarding builds genuinely useful as things you send out.
  • Can an app connect to my other tools? On Pro and above. Artifacts can call apps you have connected to Claude, with a separate artifact connectors toggle in Capabilities, and each person using the app authorises their own connections. Build and prove the simple version first, because a connected app nobody uses is just harder to maintain.
  • Are the numbers these calculators produce reliable? They are as reliable as your inputs and the logic you asked for, which is why every prompt here asks for transparent calculations you can check. None of them know market rates, none predict outcomes, and the resume checker deliberately gives you a checklist rather than an invented ATS score. Read the breakdown before you act on the total.

Want this built around your business?

We build content systems for small businesses that are not AI native. Tell us your content problem and we will come back within 48 hours with what yours would include.

← All guides