Lincei Labs

The agentic clinical research environment.

Lincei is an open-source workbench for clinical research with EHR data. Direct research agents in plain language, inspect the path from cohort logic to final figures, and keep human review in the loop.

Download Lincei Star on GitHub

Open source Release targets: macOS, Windows, and Linux

Bring your own agent Use the authenticated tools you already have.

  • CodexSupported
  • Claude CodeSupported
  • OpenCodeSupported
  • CursorEarly access
Lincei / Study workspace
Research environment
Temporary product preview of the Lincei desktop research workspace
Temporary product capture Interface preview · subject to change

01 / Auditability

The work should be visible.

A plausible answer is not enough. Lincei turns the path of a study into an inspectable record: definitions, runs, evidence, uncertainty, and human decisions.

Vitrine

The visible research surface for cards, artifacts, review gates, and provenance.

Illustrative study trail Example values, not study results
Cohort tableCurrent

Cohort definition

2,841 eligible records

mimic-ivduckdbcohort.sql
Reviewed
MeasuresAssertions

Phenotype checks

14 assertions reviewed

Accepted
Analysis viewCaptured run

Analysis run

Query, environment, and inputs captured

run_0042sha256:9b71…
Reproducible
NarrativeEvidence-linked

Methods draft

Claims linked to evidence

In review
Review gateHuman review

Interpretation decision

Decision and rationale recorded

Recorded

02 / How it works

Direct the study.
Keep the path inspectable.

  1. 01

    Ask

    Describe the research question, dataset, population, endpoint, and constraints.

  2. 02

    Run

    Agents use M4 and clinical-research skills to perform and capture the work.

  3. 03

    Inspect

    Review queries, assumptions, artifacts, figures, dependencies, and provenance.

  4. 04

    Decide

    Approve, challenge, fork, or revise before the study moves forward.

03 / EHR-native foundation

Complex data access,
made legible.

M4 gives agents structured, read-oriented access to clinical databases while preserving the dataset, backend, query, and derived inputs around each result. Rigorous definitions still matter; Lincei makes them easier to see and review.

Explore M4 on GitHub
Dataset access M4-backed
MIMIC-IV DemoBundled local sample
Bundled
MIMIC-IVCredentialed clinical database
PhysioNet
MIMIC-IV NoteCredentialed clinical notes
PhysioNet
eICUCollaborative research database
PhysioNet
LocalDuckDB
RemoteBigQuery
AccessCredentialed

04 / Clinical-research rigor

Skills that know where studies fail.

Clinician-reviewed, domain-specific checks help agents surface methodological risks early and move consequential choices into explicit human review gates.

Built for research workflows—not clinical care, diagnosis, or automated clinical decision-making. Outputs remain subject to expert review.

Clinical research pitfalls Checked before interpretation
  • 01Cohort leakage
  • 02Immortal-time bias
  • 03Survivorship bias
  • 04Missingness
  • 05Coding drift
  • 06Denominator mismatch
  • 07Inappropriate adjustment
  • 08Multiple testing
  • 09Unsupported clinical claims
Define

Protocol, cohort, estimand

Review

Assumptions and choices

Capture

Scripts, runs, artifacts

Interpret

Conservatively, with limits

05 / Open source

One open stack.
Three clear layers.

Inspect the tools, follow the design decisions, and build on an open foundation for more transparent clinical research.

01

Lincei

Research environment

Direct agents, inspect research cards, and keep study decisions in view.

Star on GitHub
02

M4

EHR data access

Give agents structured access to clinical datasets and derived tables.

Star on GitHub
03

Vitrine

Auditable surface

Keep research outputs, checkpoints, artifacts, and review decisions visible.

Star on GitHub

Project status Lincei is early. Expect rough edges; inspect the source and follow development on GitHub.

06 / Why “Lincei”

The eye of the lynx.

Lincei takes its name from the Accademia Nazionale dei Lincei, founded in Rome in 1603 by Federico Cesi and collaborators. Its lynx emblem stood for sharp, penetrating observation: studying nature through evidence rather than submission to inherited authority. Galileo joined in 1611, and the Academy later published his work on sunspots.

Lincei Labs is not formally affiliated with the Academy.

Open by design

Open clinical research
deserves open tools.

Build studies with agents while keeping the evidence, assumptions, and decisions available for human review.

Download Lincei Star Lincei on GitHub