MerchantScope
v0.2Detective mode — multi-agent investigation with cardholder context and a citable evidence trail.
Free public sources fanned out in parallel.
Triangulates the legal entity from registry + SEC + WHOIS.
Unicode confusables + Levenshtein vs. top-brand watchlist.
Cross-checks domain age, Wayback first-seen, and news volume.
Reads Reddit, news, reviews; flags confirmed-negative signals.
Fits VADER sentiment slope; flags collapses + inflections.
Names vendors, partners, parents, co-defendants from text.
Tracks cardholder ↔ merchant overlap.
Walks the entity graph for shared agents, addresses, and cert-sibling clusters — including prior flagged cases.
Lawsuit / investigation / consent-order signals.
Matches assembled findings against a fraud-archetype catalog.
Weighs every signal, pattern-matches archetypes, writes the memo.
Investigation Inputs
Only the merchant name is required. Cardholder context dramatically improves the connection check.
Paste internal transaction rows (sent to the backend and to your LLM provider)
One row per line. The backend reads the date, amount, descriptor and MCC, totals them, and hands the totals to the velocity, connection, pattern and synthesis agents. Formats it understands: ISO or US dates, dollar signs, thousands separators, tabs or runs of spaces. Rows it cannot read are counted and reported rather than dropped silently.
Run an investigation to begin
Merchant name is required. Add cardholder details to unlock Trail — the connection agent. Tap any AI mascot in the top horizontal strip to tune its model or prompt.