The deterministic verification layer beneath biomedical AI.
Paste an AI answer and get one verdict — pass / review / reject — with a receipt for every finding. Live against PubMed, Open Targets, EBI ChEMBL and ClinicalTrials.gov, no sign-up. The full depth — variant pathogenicity (gnomAD/dbNSFP), drug interactions, drug regulatory status, discovery verification and multi-source corroboration over a 493 GB datalake — runs via the CLI / MCP connector. How it works →
Finds the checkable parts — cited DOIs, gene / drug / disease claims — verifies each, and
returns one pass / review / reject verdict with a grounded-confidence score.
This is the browser demo (public APIs); the deep datalake checks run via the CLI / MCP.
Nothing is asserted without a receipt. How it works & why to trust it →
or drop your bibliography / thesis files here.bib · .tex · .md · .txt · .ris — checks every cited DOI, per file, for retractions
DOIs are extracted offline; each is resolved against PubMed and checked for
retraction, correction, and provenance. Nothing is asserted without a receipt.
Extracts the disease, drug, and gene entities and verifies each — genes against
Open Targets, drugs against ChEMBL. Nothing is asserted without a receipt.
Use this when you want to see exactly which entities were found and how each resolved;
use Verify an answer for one overall pass / review / reject on a whole answer.
(The local build uses OpenMed NER for higher recall.)
Checks the target against Open Targets — genetic association, known-drug, and
literature evidence. A weak or absent association is flagged, not hidden.
Checks a drug against ChEMBL (market withdrawal, FDA black-box warning) and a
trial against ClinicalTrials.gov (terminated / withdrawn / suspended early).
Nothing is asserted without a receipt.
Deterministic — same input, same verdictEvery verdict a receipt — no false confidenceNo false “pass” by design — the cardinal rule ↗Model-agnostic — runs inside your stack
Take it past the demo
The browser runs the public-API subset. Here's how to get the real thing into your work — all usable today.
In your AI product
Gate every biomedical answer your model or agent produces — reject or retry before a user sees it.
The demo and the open-source CLI/MCP are free. For the full-depth hosted API (variant
pathogenicity, interactions, drug regulatory, private data, SLA) or a design-partner run on your
own outputs, tell us who you are — we'll be in touch.