Building your first app

From an idea to a working app — stacks, databases, the script, and the build.

Updated 27 Sept 20263 min min read
On this page

Building your first app

Monokrew builds working software from a description in plain language. You choose how much to decide yourself and how much to leave to the agent — then watch it build, and change anything you don't like.

What you can build

Full-stack applications, websites and web pages, with authentication, a database and file storage. Every project is built for search engines from the start.

Stacks: Node.js · Next.js · NestJS · Vue · Python (web) · Django · .NET · Flutter

Databases: Postgres · Supabase · Hive · SQLite

Hive and SQLite run on the device, so a Flutter app can work without a server. Postgres and Supabase run in the cloud.

Describe your project

Give the project a name and describe what you want. Be concrete — name the kind of app, its core feature, and the look you're after. "A task tracker with a list of todos I can add, check off and delete, in a clean minimal style" gets you far closer than "a productivity app".

Then choose your stack, authentication, database, storage and theme — or let Monokrew suggest them. If you don't have a strong preference, the suggestion is usually the right starting point.

Add your designs

If you already have branding or designs, add them at the design step. You can also connect Figma and provide a URL, and the agent builds against your real designs rather than guessing at them.

Clarifications — on or off

You can let the agent ask you questions while it builds.

Turn clarifications on when you want to steer the details and you're available to answer. Turn them off when you'd rather describe what you want once and come back to a finished app — the agent has enough to proceed without you.

Neither is better. It depends on whether you want to be in the room.

Review the script

Before any code is written, Monokrew turns your description into a detailed script of what it is going to build. This is the point to change your mind cheaply.

Read it, edit it directly, or ask the AI assistant to improve it. When you're happy, start the build.

Watch it build

The agent plans, writes and assembles your app while you watch. The header shows which step it is on, how many remain, and the tokens the run is spending against your balance.

If a run runs out of tokens it pauses cleanly rather than failing — your completed steps are kept, and you can top up and carry on.

Preview and change it

The preview loads by itself when the build finishes, and it is interactive — click through it as a real user would.

From here you have three ways to change things:

  • Ask the agent. Type what you want in the prompt box — "make the header dark", "add a due date to each task". You can attach screenshots and other files.
  • Edit it yourself. Open the IDE, change the code, and the preview reloads as you type.
  • Look at your data. If your project has a database, you can view it directly.

Keep each change focused. One clear change at a time gives better results than a long list, and you can see exactly what each edit did.

Keep your code

Your source is yours throughout.

  • Push to GitHub or GitLab whenever you like
  • A secure copy is kept in the hosting cloud
  • Reload the project any time and keep editing

Next steps

When it looks and works the way you want, continue to Deploying and hosting to put it online.

Something missing or wrong? Tell us.