Compare a transparent lexical claim-checking judge against two human ratings and inspect Cohen's kappa agreement. Runs entirely in your browser.
Bundled sample claim/evidence rows (edit in src/app.ts and rebuild to try your own):
Known limitations: the judge is a simple lexical-overlap heuristic, not an LLM;
this evaluates agreement mechanics only. See the project README for the full write-up and the Python CLI version.