The Problem
30+
minutes to manually match one patient against one trial — and there are thousands recruiting at any given time
Clinicians want to refer eligible patients to trials. The time cost means most skip it entirely — so patients who could benefit never get referred.
Eligibility criteria are written in dense clinical language with complex logical conditions. Matching them against a patient chart requires careful cross-referencing, every time, for every trial. This tool reduces that to a single SOAP note paste.
What It Does
Two tools in one. Start from any note — rough or structured.
- Note Translator — paste any rough clinical note and Claude rewrites it into SOAP format; missing fields flagged automatically so the clinician knows exactly what to fill in
- Searches live ClinicalTrials.gov data — real government records, thousands of active recruiting trials, not a curated index
- Returns per-criterion verdicts with the patient’s value shown alongside each result
- UNKNOWN when chart data is missing — the tool never guesses
- Score zeroing on any FAIL — one disqualifying criterion eliminates a trial, mirroring real clinical screening
- Optional location filter narrows results to trials recruiting nearby
Outcome
LangGraph ReAct agent. Two tools. Live government data.
- Agentic orchestration — the agent decides when to call the note optimizer and when to call the screener; no fixed pipeline
- Live data — operates on ClinicalTrials.gov directly, not a cached subset; trials that closed yesterday are gone today
- UNKNOWN > guess — the system flags missing data rather than inferring it, matching the epistemic standard of actual clinical decision-making