No-code, low-code or custom development: which do you choose?

The terms first, because they get mixed up.

• No-code: you build applications from visual building blocks, without a programming language
• Low-code: a visual platform too, but one where developers write code for the parts it does not cover
• Custom development: a development team writes the application in code from the ground up
• AI coding: an AI model such as Claude or ChatGPT writes the code or builds the application from what you ask in plain language

Low-code wins in the hands of your own IT department.

If your company has developers who want to build themselves, low-code is a logical choice. The platform takes the repetitive work off their hands; your team writes the rest in code, free to change anything.

For internal tools with a limited audience, low-code is often enough. The question is who maintains it, and what the licence costs as usage grows.

No-code wins when the builder knows the process.

With no-code the line low-code draws is gone. Integrations with systems like your accounting, email automation and robot control are built from blocks too, not as exceptions written in code. Core holds 2,800 ready-made blocks for exactly that.

The ERP systems, warehouse systems and robot cells we build for customers run on precisely those same blocks. If you do hit a limit, you write one block of your own in ReactJS and use it like any other block afterwards.

Maintenance sits with the platform, not with you. We update the codebase daily, security patches included, and because your application is made of separate blocks those updates arrive without a rebuild.

You choose custom development for a reason.

If you want your own native app in Swift or Kotlin, or a product you sell as a software company, writing the code is the better choice. You need a development team for it, in house or hired, and you carry the technical debt yourself.

A middle way exists too: your processes run in Core and your app calls them through HTTP triggers, so you only write code where it adds something.

AI coding changes who builds.

With AI coding tools such as Claude Code and Codex, a model writes in minutes what takes a team days. That works with each of the three approaches, but the risk differs: in loose code, a model can forget a check somewhere or take a shortcut no review catches.

That is why we call Core an AI coding platform. The model builds there with the same blocks you do: permissions and rate limits are settings on a block, every connection to another system is a block on the canvas, and what the model makes is a draft until you publish. You keep the speed of AI, and what was built stays readable for your own people.

The comparison at a glance.

Which column fits depends on your team and your process. The Core figures are ours; project cost depends on scope.