Jev.SearchSearch That Shows
Search That Shows
Its Work.
triage → web → jev reranks every source → the model drafts → jev verifies every claim. every step typed, every step on screen.
try:
why it differs
most AI search hides the ranking. here every source gets a typed usefulness score before the model sees it, and every claim gets a verdict after.
cli
same pipeline in your terminal:
npx jevbook-bsc "your question"
streams the stages as they land.
npx jevbook-bsc "your question"
streams the stages as they land.
mcp
agents get the same via /mcp → research, verify_claim, decide, scan_token.