Clinical data infrastructure · workflow layer
Generate a specialist-ready packet
Drop in fragmented, multilingual medical records. MedLineage reconstructs the case, cites every key claim, flags missing records, and produces a specialist-ready packet for clinician review.
Note. This tool organizes medical information for visit preparation only. It does not diagnose, recommend treatments, or replace a clinician.
Upload PDFs, labs, discharge summaries, CSV exports, FHIR Bundles, or HL7 messages.
PDFs may be in any of EN / IT / FR / ES / DE; output is in the language you select.
Use de-identified records for demos. MedLineage prepares records for clinician review; it does not diagnose or recommend treatment.
To continue: At least one medical record
Build a source-cited patient record
Upload lab reports, imaging, clinical notes or discharge summaries above. MedLineage assembles them into one reviewable record — and a specialist-ready PDF packet.
Optional · longitudinal record
Longitudinal patient record
Optional. Use this when you want MedLineage to connect multiple uploads over time and organize them for specialist review.
Optional · longitudinal record
Longitudinal patient record
Optional. Use this when you want MedLineage to connect multiple uploads over time and organize them for specialist review.
Clinical record agents
What the agents did
Bounded, source-grounded agents that coordinate record-preparation work, surface gaps, and route tasks for clinician review. They do not diagnose, recommend treatment, or replace clinicians.
What agents do
Check whether expected records are present, link findings to source documents, prepare specialist packets, and route cases for human review.
What agents do not do
They do not diagnose, recommend treatment, autonomously contact patients, write to EHRs, or replace clinicians. Clinicians review the packet.
Intake agent
Organizes uploaded records: classifies documents, detects languages, and flags duplicates or unreadable files.
No records ingested yet. Upload PDFs or load a demo case to begin.
- Documents classified
- —
- Languages detected
- —
- Date range
- —
- Files needing attention
- —
Graph agent
Builds the source-grounded patient graph from the classified documents — every fact links back to the document it came from.
Patient graph not built yet. The graph is created when you finalize a packet.
- Patient graph status
- Not generated yet
- Timeline events
- —
- Extracted facts
- —
- Provenance coverage
- —
Review agent
Finds claims that need human review — low-confidence facts, disputed entries, items awaiting clinician attestation.
Review queue is empty — generate a packet to populate verifiable claims.
- Unreviewed claims
- —
- Attested by clinician
- —
- Claims in this packet
- —
- Last run
- —
MDT agent
Prepares the case for multidisciplinary discussion: source-grounded summary, missing inputs, and open questions for the team.
MDT briefing applies to oncology cases only.
- MDT briefing
- Not applicable
- Source-grounded
- —
- Open questions
- —
- Last run
- —
Surveillance agent
Surfaces factual and gap-based signals over the longitudinal record — outdated records, missing follow-up documents, contradictions.
No surveillance signals raised against this longitudinal record.
- Signals for clinician review
- —
- Outdated records
- —
- Missing follow-up
- —
- Contradictions
- —
Interop agent
Prepares safe handoffs and exports — FHIR-mappable Bundle, Data Room, signed verifiable record, write-back dry-run for clinician approval.
Exports not yet available — finalize a packet to enable handoff and signing.
- Data Room
- Not generated yet
- FHIR-mappable Bundle
- Not generated yet
- Write-back dry-run
- —
- Signed / verifiable record
- —
All agents are read-only, step-capped, and source-grounded. Approval-gated tasks require a human in the loop before they can be marked resolved.
LayoutDefault
Operational layer
Clinical Work Queue
A record-preparation worklist. It organizes cases by what record-preparation step is pending — missing documents, claims to source-review, packets ready for specialist handoff.
Not an AI triage or treatment-priority list. The queue does not rank patients for treatment or make clinical decisions. It organizes record-preparation work for human review.
Tasks are not clinical recommendations. They are record-preparation actions for human review.
Task types
- Missing records
- Claims to review
- MDT preparation
- Surveillance signals
- Approval-gated exports
Tasks will appear here after you ingest records for a case.
Open work queuePatient Graph
The Patient Graph fills in as you ingest records.
Clinician review
Attest each finding against its source. Verdicts are HMAC-signed in the Provenance Ledger.
Available after the packet is generated.
Open review consoleSurveillance signals
Deterministic detectors over the patient graph — facts and gaps only.
Surveillance signals appear here after you ingest records for a case.
Open surveillance signalsMDT briefing
Structured 12-section briefing for tumor-board discussion. Facts and gaps only — no recommendations.
The MDT briefing is available for oncology cases only.
Open MDT briefingTrust & Safety
How MedLineage stays bounded and reviewable.
- Source-grounded Every claim links back to the source document it came from.
- Clinician-review-ready Outputs are bounded review artifacts, not autonomous decisions.
- Provenance Ledger Every clinician verdict is HMAC-signed and tamper-evident.
- Best-effort de-identification Applied on de-identified surfaces. Not a certification.