Open-source clinical decision support
Cross-checks NCCN, ESMO, and drug labels for you — so you verify, not research.
One JSON patient profile in, a standard and an aggressive treatment plan out — side by side, with a citation under every recommendation. The engine reads 444 sources for you and surfaces 594 red flags worth a second look.
Rules-first, not LLM-driven. The plan cannot hallucinate a drug or a dose. The clinical decision stays with you; the engine is the second pair of eyes that read every guideline at once.
Saves the routine work
Five tabs collapse into one. The engine does the cross-checking so you can spend the time on the call, not the lookup.
The engine resolves NCCN, ESMO, drug labels, and CIViC actionability in a single pass. No more switching between guidelines, label PDFs, and biomarker databases for each case.
Every line in the plan ships with the PMID, DOI, DailyMed link, or guideline section. You verify in 10 seconds instead of running a literature search to back up each recommendation.
NCCN, ESMO, EHA, BSH, EASL, MoH, WHO, CTCAE, FDA, CIViC — refreshed on a schedule. No more monitoring every guideline update yourself to know what changed since last quarter.
Both alternatives are generated side by side every time — you don't manually construct the second option to weigh against the first. CHARTER §15.2 C6: the alternative is never buried.
Surfaces what's easy to overlook
The engine is a checklist that runs against every profile. It cannot replace clinical judgment — but it can stop a tired Friday afternoon from missing a contraindication.
Drug interactions, contraindications, organ-function thresholds, high-risk cytogenetics, bulky disease — every red flag fires automatically against the profile and surfaces in the plan for review.
CIViC actionability lookup is fusion-aware. A BRAF V600E or a BCR::ABL1 finding maps to ESCAT-tier evidence before it can quietly disappear into a long workup report.
The plan is assembled by a declarative rule engine, not by an LLM. A drug must exist in the KB with a real label entry; a dose must come from a curated Regimen YAML. No creative invention.
When the profile lacks data — missing cytogenetics, no ECOG, incomplete biomarker panel — the engine emits an Open Question instead of guessing. CHARTER §15.2 C6: anti automation-bias is mandatory.
Open-data inputs: CIViC (CC0) for biomarker actionability, ClinicalTrials.gov for trial-aware options, PubMed/PMID/DOI and DailyMed/openFDA for literature and drug-label context. No LLM chooses treatment: plans are rules-first with YAML provenance, so LLM hallucinations are excluded from the plan. The clinician remains the final authority (CHARTER §11).