Open app

For clinics today · for every healthcare agent next

The verified clinical-context layer for healthcare AI.

MedLineage reads a clinic's inbound records — PDFs, HL7 v2 messages, FHIR observations, CSV labs, in five languages — into a source-cited patient graph that clinicians attest and other AI can query.

Try a synthetic caseSee sample packet

100% provenance coverage · 1.00 extraction F1 on our published synthetic benchmark — raw JSON public, drift fails CI. JSON →

Prepares records for clinician review. Does not diagnose, does not recommend treatment, does not replace clinicians. Best-effort de-identification, not certified anonymization.

medlineage · workspacelive
Specialist-ready packet
Case A-001 · Nephrology
3 documents · 2022 → 2024 · IT · DE · EN
PDF · A4
3
Documents
3
Languages
86
Readiness
Patient graphpat_•••• · 3 docs
Active problems
Persistently flagged Creatinine across reports
What changed
Creatinine 1.2 → 1.5 → 1.8 mg/dL · upward trend
Missing for nephrology
Recent urinalysis · Kidney imaging · BP record
Source-grounded provenance
Every claim cites a doc_id + page + source language
FHIR-like Bundle
Signed Data Room
RAG over graph

Measured on synthetic complex-care demo

From fragmented records to specialist-ready intake.

Synthetic demo data
View synthetic demo packet
Source-cited transformation
10
documents in
51 pages · 5 languages · 8 doc types
fragmented medical recordsPDF · FHIR · HL7 · CSV
PDFFHIRHL7CSVXML
MedLineage
clinical record intelligence
1
specialist-ready packet
8 pages · 9 events · 37 cited findings
schema-validPDF + JSON
IT · EN · FR · ES · DE
Measured impact · synthetic demo set
51 pages · 10 documents · 5 languages8-page specialist packet · 37 cited findings · 14 completeness checks

See the full metric grid, scoreboard, and what we don’t claim on Proof & Safety.

  • Multi-format ingest
    PDFs · HL7 v2 · FHIR R4 · CSV labs — no EHR migration
  • Every claim cited
    doc · page · source language on every finding
  • Agents with human approval
    agents prepare work; clinicians sign every step
  • Five languages, mix per case
    EN · IT · FR · ES · DE — input or output

How it runs

From inbound records to cited context.

  1. 01

    Ingest

    Records arrive as PDFs, HL7 v2 messages, FHIR resources or CSV labs — in any of five languages.

  2. 02

    Structure

    Deterministic passes build the patient graph. LLMs only verbalize — they never rank, score or invent structure.

  3. 03

    Attest

    Clinicians sign verdicts into a tamper-evident provenance ledger — every claim carries its verification status.

  4. 04

    Serve

    Specialist packets, FHIR bundles, signed data rooms — and MCP tools other agents cite instead of re-parsing PDFs.

Agent-ready

Other agents don't re-parse the PDFs. They cite the graph.

An MCP surface exposes the verified patient graph to any agent runtime — eight read-only tools, every payload deterministic and citation-stamped.

MCP tools

  • ask_patient
  • get_patient_context
  • assess_readiness
  • list_evidence_gaps
  • resolve_citation
  • verify_claim
  • get_fhir_bundle
  • get_verified_context

The citation contract: every factual sentence an agent gets back carries a graph ID it can resolve — uncited sentences are stripped before they ship.

Shipped behavior, demonstrated on synthetic cases — flag-gated per deployment.

mcp · /mcp · streamable-http
> tools/call get_verified_context

{
  "patient_id": "pat_3f9c…",
  "claims": [
    {
      "citation": "obs_a41d…",
      "kind": "observation",
      "text": "Creatinine 1.8 mg/dL (2024-03-02) [H]",
      "source_document_id": "doc_77e2…",
      "verification_status": "attested"
    },
    …
  ],
  "counts": { "total": 23, "verified": 19,
              "documents": 3, "observations": 14, "events": 6 },
  "deterministic": true
}

Measured, not promised

Numbers that reproduce from a pinned benchmark.

Synthetic multilingual gold set — not a clinical study. The scoreboard is committed to the repo and pinned in CI: any drift fails the build.

1.00
Extraction F1

253 synthetic gold facts · 5 languages

100%
Provenance coverage

every recovered fact cites its source document

0 / 58
Safety redactions

no synthetic case tripped the safety net

View the raw scoreboard JSON →

Sanity Health
Pilot partner

Live clinical pilot

Our first clinical pilot is live with Sanity Health.

Sanity Health is a small private clinic in Milan with around 500 patients. The clinic pushes its inbound clinical records straight into MedLineage Connect — we turn them into a source-cited patient-context layer its clinicians can review and attest, with every finding tracing back to the document, page, and source language it came from.

  • Real inbound records

    Free-text clinical artifacts pushed straight from the clinic into MedLineage Connect.

  • Cited, not summarized

    Every finding traces back to its document, page, and source language.

  • Privacy-first

    Records flow under an opaque patient handle — best-effort de-identification, not certified anonymization.

Pilot in progress. MedLineage prepares records for clinician review — it does not diagnose, recommend treatment, or replace clinicians.

05Trust & control

Built to be verified — and honest about what we don’t claim.

Provenance

Verifiable by design

  • Every claim cites its document, page and source language
  • Every agent run is logged
  • Clinicians attest each finding — verdicts stamp the next build

Data

Privacy & control

  • Best-effort de-identification
  • Tenant-isolated
  • Dry-run by default

Interoperability

Built to integrate, not to lock in

  • Multi-format ingest: PDF · HL7 v2 · FHIR R4 · CSV
  • FHIR-like & Data Room exports
  • No EHR migration

Not a medical device. Not certified for FHIR, GDPR or EHDS conformance. No diagnostic capability or guaranteed anonymization is claimed.

Start with one packet. Grow into an agentic record-prep layer.

Generate a clinician-ready PDF today. Re-upload tomorrow to thicken the same Patient Graph. Let safe agents open the next review task and prepare a signed Data Room when you're ready to integrate — every step gated for human approval.

Open the appExplore the platform

Working with hospitals or second-opinion services? Bring 5 real cases — we run them de-identified and you sign off the packets. Run a de-identified pilot

Coordinates record-preparation work. Does not diagnose, does not recommend treatment, does not replace clinicians. No certified FHIR / GDPR / EHDS / MDR / CE conformance is claimed.