Contact

Jordi CABOT SAGRERA

Envoyer un e-mail

Publié le 23.09.2026

AI Flagship

LetzOpen: What if you could build the software you need and actually own it?

Every organisation today needs more software than it can build. A bank needs an internal tool by next quarter. A public administration needs to digitise a paper workflow. An SME needs a customer portal, and increasingly, an AI assistant inside it. The developers to build all of this are scarce and expensive, and the backlog keeps growing.

Low-code platforms promised to fix this: describe what you need and let the platform generate the application. The trouble is that most of them are proprietary. Your models, your business logic and your applications live inside someone else's product, priced on someone else's terms. Leaving means starting again. The shortcut becomes a dependency.

Luxembourg built an open-source answer. And it’s BESSER than you expect.

BESSER: model once, generate everything

BESSER is an open-source low-code platform developed by the Luxembourg Institute of Science and Technology (LIST), with the participation of the University of Luxembourg's SnT, and funded by an FNR PEARL grant. It lets any organisation, from a startup or SME to a research team or public body, go from a description of a system to working software in a fraction of the usual time.

How does it work? You describe your system specs once: its data, its behaviour, its users, its AI agents, its deployment. You do this in a browser-based editor, by drag and drop on the modeling canvas or by simply explaining what you want to build in a a plain-text description, or even from a photo of a whiteboard sketch. From that single specification, BESSER's 15+ code generators produce working software on the stacks teams already use.. Three things it is built to solve, directly:

  • It turns ideas into running software fast. Databases, REST APIs, full web applications and deployment manifests are generated from the software specifications rather than coded by hand.
  • It builds AI into applications, not just around them. BESSER uses AI to help people design what they need, and it can generate software that embeds AI, including agents thanks to the BESSER Agentic Framework.
  • It keeps the rules in the model. Business constraints are written once and then validated and enforced automatically in the generated code, so what the business specified is what the software does.

Already at work

BESSER is public catalogue, Built with BESSER, lists real applications that each started as a model.

The closest example to home is the Luxembourg AI Factory Dashboard, the collaborative dashboard of the AI Assessment Sandbox Configurator, extended from a full-stack scaffold that BESSER generated. The open-source tool Luxembourg uses to check whether AI can be trusted was itself partly built with Luxembourg's open-source low-code platform.

Ask Safely is an AI agent that queries knowledge graphs while guarding against leaking sensitive data to large language models, a direct answer to one of the main concerns organisations have about generative AI.

Adoption goes well beyond Luxembourg. BESSER is used at 27 universities, has been taught in courses across 13+ countries, and its online editor has been opened from 83 countries. The most popular use by far is generating a complete web application from a single diagram, with more than 1,800 generations to date. Behind the platform sit 59 peer-reviewed publications and two EU Horizon projects.

Open by design, sovereign by consequence

Here too, open source is the point, not a detail. BESSER is released under the MIT licence, one of the most permissive there is. Any organisation can use it, run it in its own environment, and build commercial products on top of it.

More importantly, it is built to be extended. Organisations can write their own code generators for their own technology stack, add domain-specific concepts to the modelling language, and create new diagram types in the editor. A bank, a manufacturer or a ministry can shape BESSER around its own domain instead of shaping its domain around a vendor's product. The team even publishes BESSER Skills, so AI coding assistants can help build those extensions.

“That changes the economics of low-code. With a proprietary platform, your models are locked in. With BESSER, the model is yours, the generators are inspectable, and the generated code is ordinary code you can read, maintain and take anywhere.” Jordi Cabot, Head of the Software Engineering Research Group at LIST

The project also offers support for migrating between low-code platforms, because interoperability is part of the same idea.

The development itself happens in the open: 21 contributors, 87 releases and more than 270 merged pull requests on GitHub. 

For Luxembourg, BESSER is part of a broader pattern. It turns out to be one of the several examples of public research turned into open infrastructure that others can reuse. The dashboard at the heart of the Sandbox shows how those pieces already fit together.

The BESSER editor is available at editor.besser-pearl.org and requires no installation. The source code and documentation are on GitHub. Organisations that build something with the platform can submit it to the public catalogue.

Curious what this could look like for your organisation? 

 

Comment pouvons-nous vous aider ?

By content type (optional)