Low-code
Low-code describes platforms where most of an application is assembled visually or through configuration, but where you can drop into real code for the parts that need it. The code is the escape hatch rather than the primary building material.
The distinction from no-code
No-code is aimed at people who will not write code at all, and its ceiling is whatever the platform provides. Low-code assumes a developer will be involved somewhere and provides deliberate seams for them: a custom expression in a field, a script step in a workflow, a hand-written component in a page, an API call for something the connector library does not cover.
In practice the two categories overlap heavily. Most no-code platforms have added formula languages and script steps; most low-code platforms are usable without code for simple cases. Vendors pick the label that suits their positioning.
Where it fits
The strong case is a large organisation with more internal software requests than developer capacity, and a need for those requests to be met inside a governed environment — single sign-on, audit logging, permissions and hosting handled once for everything built on the platform. That governance is often the real product. The second strong case is a workflow that is 90% standard and 10% specific to your business, where the 10% would otherwise force a full custom build.
What to watch out for
The escape hatch is narrower than it looks, and this is the misconception that hurts. "You can always write code" usually means: code in a constrained runtime, in a dialect the platform defines, in a text box inside a browser, with limited debugging, restricted library access and no straightforward way to unit test. It is enough for a calculation or a transformation. It is not enough to implement something the platform's model of the world does not accommodate.
Three related costs:
- The tooling you take for granted may not be there. Version control, code review, branching, automated tests and local development all depend on whether the platform exposes its artifacts as files. Ask specifically.
- Skills do not transfer. Deep expertise in one low-code platform is worth little on another, and hiring for it is harder than hiring for mainstream languages.
- Lock-in is by design. The visual parts of your application generally cannot be exported into anything runnable. Licence costs that grow per user or per app are the normal reason organisations start looking for the exit, and that is exactly when there is nothing to take with them.
Estimate low-code projects on the awkward 10%, not the easy 90%. That is where the schedule actually goes.
Frequently asked questions
Is low-code just no-code with a script box?
Functionally the categories overlap so much that the labels say more about marketing than architecture. The question that separates real platforms is not whether code is possible but what the code can reach: can it call any library, run outside the browser, be version controlled and be tested? If not, it is a formula field with ambitions.
Build it yourself
NorthernGo turns a plain-text description into a working web app with a database, login and a live URL. Local AI generation runs on your own GPU, is unlimited, and is free on every plan.