Summary
Linkr is an open-source collaborative data science platform designed to exploit hospital data warehouses (primarily in OMOP format, but not exclusively). It combines low-code tools with Python and R programming interfaces, making these data accessible to clinicians, data scientists and data engineers alike.
The problem Linkr tackles
Clinical Data Warehouses (CDWs) bring together huge volumes of data from patient records. Making use of them is hard for three reasons:
- Every hospital has its own data model, which hurts interoperability and research reproducibility. Common data models such as OMOP or FHIR exist, but adopting them takes effort.
- Manipulating and analysing these data usually requires skills in programming (SQL, Python, R) that clinicians often don’t have.
- Existing tools sit at two extremes: programming environments (RStudio, Jupyter) that are powerful but inaccessible to non-developers, and no-code tools (e.g. ATLAS) that are limited in what they let you extend or script.
Linkr is designed to bridge that gap: give non-developer clinicians effective graphical tools, while giving data scientists a real integrated Python + R environment.
Code environments
RStudio, Jupyter
Powerful and flexible, but inaccessible to non-developers.
Linkr’s approach
Low-code
Graphical interfaces for the clinician, a Python + R IDE for the data scientist, in one collaborative tool.
No-code tools
ATLAS (OHDSI)
Accessible to non-developers, but limited in what they let you extend or script.
Who Linkr is for
Linkr is built for an entire chain of roles working around health data:
Healthcare professionals
Explore their department’s data, build cohorts and examine patient trajectories — without writing code.
Data scientists and statisticians
An integrated IDE (Python + R) with direct data access, and the ability to create and share custom analysis plugins.
Data engineers
Schema definition, concept mapping, quality checks, transformation pipelines that run on warehouse data.
Health students
Training on clinical data analysis in a guided environment.
What makes Linkr distinctive is that these four profiles work in the same tool, on the same projects, each at their preferred level of abstraction.
Collaboration as a pillar
Linkr is first and foremost a collaborative platform. The app is organised into workspaces (shared containers at the department, hospital or network level) that contain:
- Projects (studies, analyses, dashboards).
- Plugins reusable across projects.
- A wiki for team documentation.
- Shared data and schemas.
- A Git integration that lets you version and share everything across centres.
This collaborative model follows a bottom-up approach: each team starts by solving its local problems, then, by sharing scripts, cohorts and plugins through Git, standardisation emerges gradually — rather than being defined upfront. The practical payoff: results arrive faster, and each project enriches the data for the next ones. This is the principle of a cumulative investment, where quality is built up project after project.
Effort per project
Variables to prepare, project after project, on the same warehouse
Project 4: 91% already prepared
What Linkr does
Linkr covers the whole chain, from preparing the data to reporting the results.
Prepare the data
Databases
CSV, Excel, Parquet, DuckDB; PostgreSQL and others in server mode.
Schemas
OMOP CDM, MIMIC or an in-house schema, read through presets.
Concept mapping
Local codes to SNOMED CT, LOINC, RxNorm.
Data quality
Rule sets, run and kept in history.
ETL pipelines
Convert a source to OMOP.
Data catalog
Aggregate, anonymised counts, described in Health-DCAT-AP.
Analyse and report
Cohorts
Visual criteria, generated SQL, attrition.
Patient data
One patient's journey, as a timeline and widgets.
Datasets
Wide format, one row per patient, and its analyses.
Dashboards
Tabs of widgets, with filters.
IDE
R, Python and SQL, notebooks, terminal.
Reports
Write a document from the results.
Extend and share
Plugins
Write your own analysis widgets, in R or Python.
Wiki
The team's documentation.
Import and export
A ZIP archive of any entity.
Git versioning
Link an entity to a repository, push, pull.
Community catalog
Install in one click what others published.
AI agents
An assistant that acts with your rights, through the MCP server.
What makes Linkr different
One tool for every profile. Linkr doesn’t force clinicians to learn Python, nor does it force data scientists to work inside a rigid no-code application. Both work side by side, on the same projects.
Low-code by default, code when needed. Graphical interfaces cover most day-to-day needs. A full IDE is available the moment a use case calls for it.
OMOP first, but not only. OMOP is the best-supported format. Schema presets already let you work with custom hospital schemas, and will eventually support FHIR and other models.
Open source and open. Licensed under AGPL-3.0 (the export format and client libraries are under Apache-2.0). The code is publicly available and the product can be self-hosted. Everything produced inside the platform (scripts, plugins, studies) can be shared via Git.
Multilingual and clinician-friendly. The interface is available in French and English, and can be extended to other languages. The wording stays accessible to healthcare professionals who are not developers.
Next steps
- Understand deployment modes (browser vs. server) and what each mode enables.
- Launch Linkr online in two minutes: quickstart.
- Install Linkr locally: local install.