BESSER
besser-pearl.org
What it is
BESSER is an open-source low-code platform for building smart applications by modeling once and generating APIs, databases, agents, and deployments. Its 15+ generators cover databases, REST APIs, web apps, AI agents, and deployment manifests. Built-in modeling includes class diagrams, state machines, object diagrams, agents, and deployment specifications. The GUI Creator deterministically generates complete React frontends from platform-independent GUI models, and AI-assisted modeling supports software with embedded AI components. BESSER connects through an MCP server to Claude Desktop, Cursor, and other MCP-aware clients; BESSER Skills also integrate with Claude Code, Cursor, and Copilot for model and code-generation workflows. Generated Docker components can be used to deploy a running application from a model. The project is MIT-licensed and free, with no freemium tier or hidden fees. It runs on the web, Linux, macOS, and Windows. The Python package installs with pip, and the full codebase can run locally for extensions and generators. It is tested on Python 3.10, 3.11, and 3.12; Python 3.13 is not supported yet. Support includes an on-page assistant, email, and weekly GitHub bug triage.
Spec sheet
- What it does
- BESSER is an open-source low-code platform that turns models into APIs, databases, agents and deployment artifacts.Read 15 Oct 2026 · besser-pearl.org
- Audience
- BESSER is built by a research-led team at LIST and SnT / University of Luxembourg for low-code and AI-assisted software engineering.Read 1 Oct 2026 · besser-pearl.org
- Deployment
- Generated Docker components can be used to automatically deploy a running application from a model.Read 1 Oct 2026 · besser-pearl.org
- Free plan
- YesRead 22 Sep 2026 · besser-pearl.org
- Integrations
- Documented generation targets include Python, FastAPI, Django, SQLAlchemy, SQL, Java, React, PyTorch and TensorFlow.Read 15 Oct 2026 · besser-pearl.org
- Security
- For the Spec-Driven Agent free tier, no API key is required, and any user-supplied key is never stored.Read 1 Oct 2026 · github.com
- Support
- The project offers an on-page assistant without an account, email support, and GitHub bug triage weekly.Read 1 Oct 2026 · besser-pearl.org
- License
- BESSER is open source under the MIT license.Read 1 Oct 2026 · besser-pearl.org
- Headquarters
- Belval, Esch-sur-Alzette, LuxembourgRead 28 Sep 2026 · besser-pearl.org
- Agent framework
- The BESSER Agentic Framework is LLM-agnostic and supports tools, skills, workspaces, RAG, SQL querying, web crawling and multi-agent A2A communication.Read 15 Oct 2026 · besser-pearl.org
- AI coding assistants
- BESSER Skills integrate with Claude Code, Cursor, and Copilot for creating models and code-generation workflows.Read 1 Oct 2026 · besser-pearl.org
- AI modeling
- Users can start from natural-language descriptions or images and use AI-assisted modeling.Read 15 Oct 2026 · besser-pearl.org
- Backend generator
- The backend generator combines FastAPI, SQLAlchemy and Pydantic and provides CRUD, pagination, search and relationship endpoints.Read 15 Oct 2026 · besser.readthedocs.io
- BPMN
- BESSER supports BPMN 2.0.2 workflow modeling and BPMN XML import and export.Read 15 Oct 2026 · besser-pearl.org
- Business rules
- Its B-OCL tooling parses, validates and propagates invariants and pre/postconditions into generated code.Read 15 Oct 2026 · besser-pearl.org
- Company facts
- LinkedIn lists BESSER-PEARL as a Luxembourg nonprofit with 11–50 employees, founded in 2023 and headquartered in Esch-sur-Alzette.Read 15 Oct 2026 · lu.linkedin.com
- Documentation
- ReadTheDocs covers installation, tutorials, B-UML, generators, API reference, troubleshooting, contributor setup and release notes.Read 15 Oct 2026 · besser.readthedocs.io
- Funding
- Luxembourg's FNR says Jordi Cabot's BESSER PEARL chair received €3 million and was established with SnT collaboration.Read 15 Oct 2026 · fnr.lu
- Generators
- BESSER advertises more than 15 generators for databases, REST APIs, web applications, AI agents and deployment manifests.Read 15 Oct 2026 · besser-pearl.org
- GUI Creator
- GUI Creator models screens, navigation and data binding visually or in natural language and deterministically generates React frontends.Read 15 Oct 2026 · besser-pearl.org
- GUI generation
- BESSER GUI Creator generates complete React frontends deterministically from platform-independent GUI models.Read 1 Oct 2026 · besser-pearl.org
- Installation
- The Python package is installed with pip, and the full codebase can be cloned and run locally for extensions and generators.Read 1 Oct 2026 · besser.readthedocs.io
- Licence
- The core BESSER project is released under the MIT licence.Read 15 Oct 2026 · github.com
- Limit: behavior coding
- The same evaluation says generated output is largely static and interactive behavior still requires Python and Django ORM code.Read 15 Oct 2026 · link.springer.com
- Limit: incomplete file support
- A Software and Systems Modeling evaluation reports that the Django generator lacked complete file-field handling and only partially persisted files.Read 15 Oct 2026 · link.springer.com
- Limit: UML compliance
- A 2024 independent article characterizes B-UML as UML-inspired rather than fully UML-compliant and says early versions were mainly static.Read 15 Oct 2026 · modeling-languages.com
- MCP integration
- The editor includes an MCP server intended for clients such as Claude Desktop and Cursor.Read 15 Oct 2026 · besser-pearl.org
- Model types
- The platform includes class, object and state-machine diagrams plus agent and deployment specifications.Read 15 Oct 2026 · besser-pearl.org
- Neural networks
- The neural-network metamodel generates runnable PyTorch or TensorFlow code, including training, evaluation and saving logic.Read 15 Oct 2026 · besser-pearl.org
- Online editor
- The hosted editor lets users model, generate and deploy in a browser without installing software.Read 15 Oct 2026 · besser-pearl.org
- Personalization
- Structured user profiles can drive adaptations of presentation, modality, content and agent behavior.Read 15 Oct 2026 · besser-pearl.org
- Praise: extensibility
- An independent research paper describes BESSER's notations and generators as extensible and open to community contributions.Read 15 Oct 2026 · arxiv.org
- Praise: openness
- An independent 2026 paper highlights the freely accessible editor, transparency and extensibility as responses to commercial low-code lock-in.Read 15 Oct 2026 · arxiv.org
- Public API
- The web-editor backend exposes FastAPI routers for generation, model conversion, image or CSV import, OCL validation, GitHub deployment and agent simulation.Read 15 Oct 2026 · besser.readthedocs.io
- REST API
- The REST generator creates FastAPI services with selectable GET, POST, PUT, PATCH and DELETE methods.Read 15 Oct 2026 · besser.readthedocs.io
- Runtime limit
- BESSER is tested on Python 3.10, 3.11, and 3.12; Python 3.13 is not supported yet.Read 1 Oct 2026 · besser.readthedocs.io
- SDK
- BESSER ships a Python library and Python SDK built around the B-UML metamodel and generators.Read 15 Oct 2026 · github.com
- Security change
- Release 7.16.1 fixed a validation endpoint issue that could expose unexpected server exceptions and stack details.Read 15 Oct 2026 · github.com
- Self-hosting
- Users can clone the repository, install dependencies, run Docker Compose and start the backend and frontend locally.Read 15 Oct 2026 · besser.readthedocs.io
- Simulator limitation
- The live agent simulator requires a separate service token and sandboxing, defaults to one session per user and rejects custom Python code unless enabled.Read 15 Oct 2026 · github.com
- Spec-driven API
- The documented spec-driven service offers generation, preview, run-status, event-stream, cancellation, download, GitHub push and GitHub import endpoints.Read 15 Oct 2026 · besser.readthedocs.io
- System requirements
- Current documentation tests Python 3.11 and 3.12; local editor development additionally needs Git, Docker and Node.js 20 or newer.Read 15 Oct 2026 · besser.readthedocs.io
Plans and prices
Strong points and limits
Strong points
- 15+ generators for application components and deployments.
- GUI Creator generates React frontends from GUI models.
- MIT-licensed and free, without a freemium tier.
- Works with multiple MCP clients and coding assistants.
- Support includes an on-page assistant and email.
Limits
- Python 3.13 is not supported yet.
- Python support is listed for versions 3.10 through 3.12.
Who it suits
BESSER suits developers and teams interested in low-code or AI-assisted software engineering through models and generated artifacts. Its local codebase and extension options may also fit users who want to work with its generators.
BESSER in depth
Read the full write-up
Overview
BESSER is an open-source low-code platform that turns software models into generated application components. It is best suited to software teams that want a model to drive both conventional code generation and AI-assisted workflows; its breadth is compelling, provided the team can work within its supported Python versions.
Made by a research-led team at LIST and SnT / University of Luxembourg, BESSER was founded in 2023 and is licensed under MIT. Its model repository and team collaboration support shared work, while its focus on generated outputs makes it more than a diagramming tool.
For readers comparing categories, BESSER sits among UML Modeling Software and No-code app builders, though its distinctive emphasis is model-driven software generation.
Key features
Models that produce software
BESSER offers 15+ generators for databases, REST APIs, web apps, AI agents, and deployment manifests. Generated Docker components can help turn a model into a running application, giving teams a route from specification toward deployment. That range is useful when a project needs several artifacts to follow one model; teams looking only to draw diagrams may find the generation-oriented scope unnecessary.
Modeling and GUI generation
Built-in modeling includes class diagrams, object diagrams, state machines, agents, and deployment specifications. The GUI Creator generates complete React frontends deterministically from platform-independent GUI models. This gives interface work a defined model-to-output path, rather than requiring the model to remain documentation beside separately built screens.
AI-assisted workflows
BESSER supports modeling and generating software with embedded AI components. Its MCP server connects to Claude Desktop, Cursor, and other MCP-aware agentic clients, while BESSER Skills integrate with Claude Code, Cursor, and Copilot for model creation and code-generation workflows. These connections make it a fit for teams already using those tools, but the core value remains the ability to generate from models.
Installation and compatibility
The Python package installs with pip, and teams can clone and run the full codebase locally to extend generators. BESSER is tested on Python 3.10, 3.11, and 3.12; Python 3.13 is not yet supported. That restriction is a practical adoption check for teams standardizing on the newest Python release.
Pricing
BESSER is free and open source under the MIT license. The BESSER plan costs nothing, with no freemium tier or hidden fees. There is no paid plan to unlock for more capability, making it straightforward for teams to evaluate or use without a software subscription.
For the Spec-Driven Agent free tier, no API key is required, and any key supplied by a user is never stored. Support includes an on-page assistant without an account, email support, and weekly GitHub bug triage.
Platforms
BESSER supports web, Linux, macOS, Windows, API, and self-hosted use. The Python package and locally runnable codebase suit teams that want to extend the platform or its generators, while the web and desktop operating-system support offers additional ways to work with it.
Who it's for
BESSER is a strong fit for teams building software from structured models, especially those that want generated APIs, databases, interfaces, AI agents, or deployment artifacts to share a modeling foundation. Its open-source license and local codebase also suit organizations that want to inspect or extend generators. It is less suitable for teams that need Python 3.13 compatibility or want a narrow, conventional UML editor without a broader generation workflow.
Pros and cons
Broad model-to-artifact coverage: 15+ generators span databases, APIs, web apps, agents, and deployment manifests, reducing the gap between a model and implementation materials.
Deterministic React GUI generation: platform-independent GUI models can produce complete React frontends, giving teams a repeatable interface workflow.
Open source and extensible: MIT licensing, a free plan, and a locally runnable codebase let teams use and extend the platform without a paid tier.
Python 3.13 is unsupported: teams already standardizing on that version will need to use a supported runtime or choose another tool.
Alternatives
ArgoUML is another free, open-source option for Linux, macOS, and Windows, licensed under EPL 1.0 and requiring Java 5 or later; choose it if that UML-editor setup better fits your environment.
BOUML is free to use on Linux, macOS, and Windows; consider it if you want a free desktop modeling alternative.
BridgePoint offers a free, Apache 2.0-licensed package with an editor, verifier, and model compilers on Linux, macOS, and Windows; choose it if those components match your modeling workflow.
Gaphor is a free, Apache 2-licensed option for Linux, macOS, and Windows; consider it as another open-source desktop modeling tool.
Horusec is a free, Apache 2.0-licensed option with CLI and platform components across API, extension, Linux, macOS, self-hosted, web, and Windows; choose it if that combination suits your setup.
Violet UML Editor is a completely free cross-platform UML editor for Linux, macOS, web, and Windows; it is an alternative if UML editing is the priority.
Visual Paradigm has a free web plan restricted to non-commercial use and a Starter plan at $4 a month, billed annually; consider it if a web-based tool fits and the free plan's use restriction works for you.
Astah Professional offers an individual perpetual license with pricing shown as starting at $11.99; consider it if you prefer that licensing approach.
Verdict
Choose BESSER if your team wants one model to feed a wide range of generated software artifacts and AI-assisted workflows, with open-source access to extend the toolchain. Look elsewhere if Python 3.13 support is essential or your needs stop at UML editing: BESSER’s defining advantage is the model-to-software pipeline, not diagramming alone.
Verdict
BESSER links modeling to a broad set of generated software artifacts and supports several AI-assisted workflows. Check the Python version requirement before adopting its package or local codebase.
Where Specifiction places it
- UML Modeling Software#1 of 269.0
- No-code app builders#73 of 1778.3