agents

Official reference rate vs USD for one of KES, NGN, GHS, TZS, UGX, ZAR, EGP, RWF, via Frankfurter (frankfurter.dev). Frankfurter blends published central-bank reference rates; the number is that blend, not a single central-bank fixing. Each response names the source and carries as_of (Frankfurter's rate date) and fetched_at.

All eight official reference rates vs USD in one payload, via Frankfurter (frankfurter.dev). Same blend and the same as_of / source fields as the single-pair shelf.

Street rate for USD-NGN or USD-GHS: the median VWAP of Bybit's public USDT/fiat P2P order book, divided by a USDT/USD spot basis, plus the Frankfurter official reference rate and the spread. OKX is archived and does not set the price. Not a central-bank rate and not a bureau de change quote.

Hourly street mid/buy/sell, spread and depth for USD-NGN or USD-GHS from Ni Biashara's own archive of the Bybit-derived quote. Each point is the hour the book was captured.

Same Bybit-derived street series as the 24h shelf, seven days deep, from Ni Biashara's own archive.

Every Frankfurter official pair plus both Bybit-derived street quotes (USD-NGN and USD-GHS) and their spreads, each row carrying its source and time. One call for a morning pricing check.

Live federal FMCSA record for a US motor carrier: legal name, operating status (allowed to operate), authority, safety rating, fleet size, crash and inspection summary — with plain risk flags. The check a freight-broker or dispatch agent runs before tendering a load.

The carrier's Behavior Analysis & Safety Improvement Category (BASIC) measures from FMCSA's Safety Measurement System — unsafe driving, hours-of-service, vehicle maintenance and the rest — with intervention-threshold flags. The depth check behind carrier-authority-check: run both before tendering.

Is this broker real and allowed to arrange freight? Live FMCSA record by MC/docket number with broker-authority status and plain flags. The check a carrier or factoring agent runs before accepting a load from an unknown broker.

One call before you tender: FMCSA authority and SMS BASICs for the carrier (USDOT or MC), FMCSA broker authority for the broker on the load (broker_mc), plus an OFAC name screen when you pass a legal name. Verdict is TENDER, HOLD or REJECT from those public fields only — no invented fraud score. Cites FMCSA QCMobile and OFAC as_of.

Screen a person or company name against the U.S. Treasury OFAC Specially Designated Nationals and Consolidated lists (~40k names + aliases; every result carries as_of and checked_at so you can see how current the list is). Deterministic normalized-name and token matching with a CLEAR / REVIEW / HIT verdict — the check a payments, marketplace or KYC agent runs on every new counterparty.

Exact-match a blockchain address against Digital Currency Address identifiers published in the OFAC SDN XML (the same Treasury list we ingest for name screens). CLEAR or HIT. No third-party chain intel, no clustering, no risk score.

One payment, 110 credits ($1.10 of calls for $0.99). 1 credit = $0.01 of list price, so a $0.10 carrier check costs 10 credits and a $0.02 sanctions screen costs 2. Valid on every data shelf priced ≤ $0.25 — FX, carrier/broker/sanctions checks, Texas CE, the load-vet pack and instant Swahili localization. Redeem with the X-SHELF-PASS header — no wallet and no payment round-trip per call, which is the point for agents that check every load or counterparty.

Deterministic TDI rules: exactly what a Texas insurance licensee needs to stay compliant — CE hours, classroom/ethics requirements, renewal fees — by license line and years licensed. Rules-engine output, not an LLM guess.

Deterministic TREC rules: continuing-education hours, renewal fees and active-license requirements for Texas real-estate licensees, by license type. Rules-engine output, not an LLM guess.

Machine localization (open translation model, run in-process) delivered in the paid response — no ticket, no email, no wait. Right for UI strings, alt text, short product copy an agent needs localized mid-run. Machine-grade output, honestly labeled: for native-quality human-reviewed copy up to 500 words, buy swahili-localization-500w.