Build with AI. Audit in Excel.

Molnify App Builder: from a description or an Excel model to a working web app

Molnify App Builder is an AI that builds working web apps. You describe the app you want, or upload an Excel model you already use, and it builds the interface, the database and the logic. The calculations stay as spreadsheet formulas, so you can open the file and read them.

Start 14-day free trial

14 days free · 200 free credits · No credit card

The Molnify App Builder robot mascot
  • 17,653 apps built
  • 15,796 registered users
  • Ten years in production
  • Apps from 2017 still running
  • Three of four AI models run inside the EU

Figures as of August 2026.

Build with AI.

The Super Consultants Quote Builder app built with Molnify App Builder, dark interface with orange accents, line items for consulting services, a grand total of $37,375.00 and a PDF quotation preview

Audit in Excel.

The Excel model behind the Super Consultants Quote Builder, open in Excel: quote number, client fields, services, hours and rates, next to the subtotal, VAT and grand total cells matching the app, with an ordinary Excel formula in the formula bar
The formula bar reads ="$"&TEXT(B10*B11,"#,##0.00"): hours times rate, formatted as dollars. Ordinary Excel.
Open the live app

Trusted by teams at

VattenfallTelia CompanyIf SkadeförsäkringarHertzAmazon Prime VideoOffice ManagementFocus NeoIndigrowVattenfallTelia CompanyIf SkadeförsäkringarHertzAmazon Prime VideoOffice ManagementFocus NeoIndigrow

“The App Builder has worked really well! It was easy to sit with a preview and actually see that it works, not to mention having someone who helps and checks that what you are building will actually work. Very pleased.”

Christian Schanner, Eccoci

Chat with the App Builder

Write text, get an app you can audit and validate

Pick your preferred AI engine: Claude, Gemini, GLM or Kimi. Then describe your app to the Molnify App Builder by simply writing to it. It might ask a few follow-up questions, then it goes off and builds. Once the app is ready it shows you the result on the right, tests it, and puts on the finishing touches. You can keep iterating, ask for more functionality, or take the app live with Publish to production. And you can always download the Excel file that holds your entire app: the App Builder keeps every equation and formula as ordinary Excel formulas, so you can audit the app in Excel without understanding code. Basic Excel skills are enough. We call it vibe coding without the black box.

A Molnify App Builder session: the chat on the left describes the product catalogue app being built and lists the spreadsheet formulas behind it, while the finished application runs live in the preview on the right

Upload your own Excel file

Upload the Excel model you already have

You can also upload your own spreadsheet, and the Molnify App Builder turns it into a web app. Type something as simple as Turn this into an app and it will, or be specific and ask for styling, extra features, a database or a connection to an external API. Unless you tell it otherwise, the App Builder keeps your file, equations and formulas exactly as they are, and adds a new sheet with cell references to your inputs and outputs. The finished application stands on a file you already know, so changing a text, a formula or a price in the pricing tab is something you can do yourself in Excel. Or ask the App Builder to change it for you. It is up to you.

The Molnify App Builder with CO2 Footprint 2025 FINAL v3.xlsx uploaded: a magnified view of the attached file chip and the Upload xlsx button, above a one-paragraph prompt asking for an app that keeps all calculations in the spreadsheet exactly as they are

The results

Here is what that prompt produced: a trusted Excel file for carbon emissions became a modern web app with a custom frontend and PDF reporting. What makes this different from other AI code builders? Every formula, emission factor and the baseline are identical. The only changes were the wiring that reads the app’s inputs. Want to be sure the AI did not touch your calculations? Check the Excel file. It contains the exact formulas and equations the finished app runs on.

The finished CO2 Footprint 2025 web app built from the uploaded Excel model, wearing a custom auditor look: total emissions 283.4 tonnes CO2e, 10.5 percent below the 2024 baseline, scope 1, 2 and 3 cards, reduction pace gauges and draggable activity bars for electricity, district heating and natural gasOpen the live app in a new tab →

The build shipped as a headless app with a custom frontend: ask for that in your prompt and the App Builder replaces Molnify’s standard frontend with one built to your description, so the app looks and works exactly the way you want. If you would rather not bother with design instructions, that is fine too. It simply ships on Molnify’s standard frontend.

Start 14-day free trial

Vibe coding without the black box

Why does that hold? Because a Molnify app has always been an Excel file. For ten years the recipe has been the same: colour the input cells green, the outputs red or blue, put JavaScript, CSS and SQL in purple metadata cells, and upload. The file becomes the app deterministically. No AI interprets or guesses at runtime; the only AI in a finished app is an AI action you placed there yourself. The file = the app. The App Builder was taught to build the manual way: it writes that same file and uploads it like anyone else. It happens to be an Excel expert, a JavaScript expert, a designer in CSS and HTML, fluent in SQL, and it knows every Molnify feature: databases, autofills, external API actions, PDF templates, email actions. But what it produces is still a file that becomes an app. Nothing changed there. That is why there is no black box: what the AI writes, Excel can open.

Same AI, two outcomes

Eighty-five per cent of developers say AI has moved the bottleneck from writing code to reviewing and validating it. A fair problem when code is the product; a different one when the thing being checked is the price you quoted a customer. Here, you do not need to be a programmer to see that something needs fixing, and you do not need one to fix it.

Question Molnify App Builder AI code builders
What you get out A working app plus the Excel file A working app plus a codebase
Where the logic lives In spreadsheet formulas In generated code
Can you audit it Yes. Open the file. Only if you can read code.
Can you fix it yourself Yes. Edit the formula. You ask the AI again.
Where the AI runs Three of four models inside the EU Typically US cloud
If the tool shuts down Your file still opens in Excel Your code needs a new home

Start 14-day free trial

Both ways of building end in the same thing: an app standing on an Excel file. And the file is the difference. Most tools in this category read a spreadsheet as a data table. Here the file is the build instruction: real Excel formulas, more than 200 functions, plus the JavaScript, CSS, SQL and actions that power the app, all in the same file. Change the app by telling the App Builder, or open the file, edit, upload again. An AI that builds, a file anyone can audit, models inside the EU, one price for any crowd. Each of those exists somewhere else. The combination does not.

“I’m a marketer and an Excel user, not a developer. My Excel simulator has evolved over the years and reached a rather high level of complexity. Mathematical formulas are therefore the core of my professional activity, and it’s essential that they remain the basis of any modification. With Molnify and their powerful AI development tool App Builder, I was able to develop key pieces of code completely independently, without the need for external developers, and in a very short time. Honestly, the most amazing thing was the ease with which I interacted with the App Builder AI agent. It really feels like having a developer available 24/7 who instantly understands what you intend to do. I think App Builder is a real game changer.”

Federico Duregon, FVSyncro, Italy

You pick the AI, and you can see which one

ClaudeOutside the EU

The default. Most proven for Molnify apps, needs the least instruction to get it right.

GeminiInside the EU

Google’s own European service. Cheaper, asks more questions.

GLMInside the EU

Open weight on European hardware, at a lower cost than the frontier models.

KimiInside the EU

Open weight on European hardware. Long context, good at filling in what you left unsaid.

Claude is the default; if EU processing is a requirement, pick Gemini, GLM or Kimi. You choose per build, with more on the way.

What can it build?

If the logic fits a spreadsheet, it fits here: ROI and business case models, insurance and risk calculators, pricing and quoting tools, engineering calculations and inspection apps for the field.

Nor does it stop at single apps. Molnify runs entire business systems: six or seven apps composed into one, on their own database tables, calling external APIs, sending email, generating PDF reports and running AI actions.

From zero to building in a minute

  1. Create a free account. Email, password, one click in the confirmation mail. No credit card.
  2. Describe your app, or upload your Excel file. The first prompt works right away, on free starting credits.
  3. Watch it build. Nothing goes live until you press Publish.

Create an app

Run the App Builder inside your own AI assistant

The App Builder skill is free and open source: instructions that teach your own AI coding assistant the rules Molnify builds on, from the cell colours to the UI commands and the metadata.

npx skills add https://app.molnify.com

It runs on your machine, inside Claude Code, Codex, Cursor, Gemini CLI or Copilot. You pay whoever runs your assistant; the skill costs nothing, and the source sits on GitHub with a checksum you can verify.

The builder, in your browser

Nothing to install, so it also works on a locked-down work laptop. It runs on Molnify Credits, you pick the model per build, and it only ever touches the app you are building.

The skill, in your assistant

Your agent works in your own files, which is exactly why it reaches further: headless apps, a frontend of your own design. That reach earns the care you would give any program that can change your files. Give it its own folder, and read the permission prompts.

If coding assistants are new to you, start in the browser. Nothing is lost by taking up the skill later. If you already live in one, it hands you the whole surface.

The skill guide · Source on GitHub

A phone running the live Datacenter ROI calculator built with the Molnify App Builder: inference cost comparison, latency, rate and IT load cards, a deployment region picker and a dotted globe visualisation

What it costs

Two costs, and only one of them keeps running.

Building costs Molnify Credits: everyone gets 200 free credits to start with, and a small app, like an ROI app, usually takes between 100 and 200 credits depending on complexity. After that, from 2.20 EUR per 100 credits.

Running costs from 99 EUR per app per month, with unlimited users. Two hundred people on your app costs the same as two: you know what an app costs, all the time, and the price does not grow with your team. Your calculations cost no credits, ever, because the finished app is not an AI answering questions. It is your model, calculating.

Your first 14 days are free. Build as many apps as you want; when the trial ends, you pay only for the ones you keep online.

You are never locked in behind the credit meter. Download the xlsx, fix the formula in Excel, and upload it again. That costs zero credits.

Not every app belongs here. If yours is not built around logic someone must be able to trust (a content site, a game, a storefront), an AI code builder is the better tool. Molnify is for the apps where someone will ask why the number is what it is.

“We turned our Excel file used for quotes into a web app using Molnify! We feel like we’ve gone from a clunky Excel file to a state-of-the-art sales tool that’s accessible to our sellers on any device. The fact that sellers can also save quotes privately in the app is also a big advantage.”

Niklas Akerö, Product Manager IT, Office Management
Start 14-day free trialOr book a demo

Common questions

What does it cost?
Building costs Molnify Credits: everyone gets 200 free credits to start with, and a small app, like an ROI app, usually takes between 100 and 200 credits depending on complexity. After that, bundles start at 2.20 EUR per 100 credits and never expire. Credit rates per AI model can change if our providers change theirs, in both directions, always from your next billing period and with advance notice by email. Running costs from 99 EUR per app per month with unlimited users, and app prices have not increased in ten years. Some capabilities are add-ons with their own monthly price, and an app that itself calls an AI while running, through the AI Prompt action, draws credits when it does. Your calculations never draw credits. The full price list is on the pricing page, and your first 14 days are free, including the apps you build in them.
Do I need to know Excel?
No. You can describe the app in plain language and the App Builder writes the model, the interface and the database. Excel knowledge matters at exactly one point: checking. The logic ends up as spreadsheet formulas, so anyone who can read =SUM or =VLOOKUP can open the file and see why a number came out the way it did. That is a far lower bar than reading program code. If you would rather build the traditional way, that path is unchanged: build the model yourself in Excel, mark it up with Molnify’s colour conventions, and upload the file.
Can the AI touch my formulas?
Only when you ask it to. Upload an existing model and the formulas carry over as they are: Molnify reads the values, formulas and colours already in the file, and the app layer can sit on a separate sheet so your original sheets stay untouched. When you ask the App Builder to change something, it edits the file, and the file stays readable: download it and compare before and after whenever you want certainty. The people using your finished app can never touch a formula at all. They see inputs and results, not the model, so nobody can overwrite a cell by accident and quietly break the maths.
Does it support VBA and macros?
Molnify apps do not run VBA or macros. The logic of a Molnify app lives in spreadsheet formulas, actions, autofills and JavaScript instead, which is what keeps it readable in Excel. But a macro-heavy workbook is not a dead end: upload it and tell the App Builder to convert, with a prompt like Turn all the VBA and macros in this file into Molnify functions, actions, JavaScript and autofills. It reads what the macros do and rebuilds that behaviour from Molnify’s own building blocks, and the converted logic ends up in a file you can open and check.
Can a non-programmer check what the AI built?
On most AI platforms, no. Verifying the business logic behind an AI-built app means reading the generated code or the database policies, so a non-programmer is left trusting the AI’s own summary or paying for an expert review. In Molnify the logic is spreadsheet formulas. The bar for checking a price rule or a threshold is opening the file and reading =IF or =VLOOKUP, which is Excel literacy rather than a code review, and most finance and operations teams already have that skill in the room. That is the design decision the whole platform is built on: the AI writes the model quickly, and a person who never writes code can still see exactly what it wrote.
How often do spreadsheet models actually contain errors?
More often than anyone likes to hear. Audits of business-critical spreadsheets keep finding errors in roughly nine out of ten of them, a figure that has held across decades of research. The costs are documented too: a copy-paste error once shifted the rows in a company’s bidding spreadsheet, so it bought contracts it did not need at prices it did not intend, and the mistake cost 24 million dollars, about a tenth of that year’s profit by contemporary accounts. The formula was not complicated. Nobody had opened it. That is the case for building on a model someone can open: the question was never whether your model has a mistake somewhere. It is whether anyone can find it.
What if the AI gets a formula wrong?
Then you can catch it, which is the point of the whole architecture. Every build runs through Molnify’s validator, which flags common mistakes on the spot. For everything else the logic is ordinary spreadsheet formulas: download the file, open the cell behind the number you doubt, and follow it through the cells it depends on until you find the step that went wrong. Fix it yourself in Excel and upload the file again, or tell the App Builder what you found and let it make the correction. Any AI can be confidently wrong. The difference here is that checking the work takes spreadsheet literacy rather than a code review.
What happens when our price list changes in March?
That cycle is the everyday life of these apps, and you choose the route each time. If you are comfortable in Excel, download the file, change the rates and upload it again. That involves no AI and costs no credits. Or tell the App Builder what changed and let it edit the model for you. Either way the app is rebuilt from a file you can open, so before your team quotes with the new version you can download it and read exactly what changed. And the version you had before is never gone: keep the downloaded file, and you can always upload it again.
Which AI models can I build with?
Four model families, each in more than one size: Claude (Haiku, Sonnet and Opus), Gemini (Flash and Pro), GLM and Kimi, the last two also in fast Instant variants. Claude is the default and the most proven for Molnify apps; Gemini, GLM and Kimi run inside the EU. You pick the model in the same window where you write your prompt, and you can switch between builds. Different models draw credits at different rates, so a light model stretches your free credits further on simple work, while the frontier models earn their keep on complex apps. The full library, with example uses for every model, sits in the AI Manager.
Where does my file end up?
On Molnify’s platform, which runs on Google Cloud in Western European data centres. It stays a file: download it from your account at any time, unchanged, and it opens in Excel like any other workbook. The AI models you build with are a separate matter, and you pick the model before the first message. GLM and Kimi are open weight models that originally came out of Chinese labs and run on European hardware here. Gemini is Google’s own European service. Claude runs outside the EU. None of them changes where your file lives, and Molnify’s standard agreement commits that your data is not used to train AI models. While you build, use sample rows rather than real records; point the finished app at your real data once it runs on Molnify.
Can I build and test without my users noticing?
Yes. Every build previews in the Lab, a private staging copy of the app that lives on its own address. You can build, break and refine there for as long as you like, and nothing your users can reach changes until you press Publish. The Lab stays with the app after it goes live, so the next change follows the same route: test privately first, publish when you are ready. And if the app replaces a spreadsheet your team works in today, nothing stops you from running the two side by side until you trust it.
What happens if Molnify disappears?
An honest answer has two halves. The app would stop: the interface, the database and the logins run on our platform, like any hosted service. Your model would not: it is an Excel file you can download today and keep, and the logic the app ran on lives in that file as ordinary formulas. You would lose the tool, not the maths, and the standard agreement includes data export on exit. The risk itself has a track record to weigh: Molnify has run for ten years, app prices unchanged the whole time, and apps built in 2017 are still in production today.
Start 14-day free trialOr book a demo

Scroll to Top