Why do two papers disagree? Compare them, isolate the methodological differences that drive the conflict, and verify every claim against the source.
Examples run on cached PubMed abstracts — no API key needed.
Fetches from PubMed and extracts with Claude. Needs the server running in live mode.
StudyDiff grounds every claim with OpenGATE — deterministic, no LLM-as-judge. Fields the source can't support are shown as not reported, never guessed.