Every check your agent can run before it acts — grouped by job, computed
from official sources, with provenance in
every response.
Paid per call in stablecoin — no account, no API key.
Most endpoints include one free call per day — try before you pay.
Before your agent commits — every check in one call
Composite pre-flights that run a whole family of checks in one call and return a single verdict you can recompute — a check that did not happen never passes, and every block carries its own provenance.
/preflight/payment
GET /preflight/payment
$0.25
Run every check on a bank transfer before your agent executes it, in one call: IBAN validity, bank and BIC from national registries, SEPA scheme reachability from the EPC register, sanctions screening of the bank and of its jurisdiction (OFAC, EU, UN, FATF), and screening of the beneficiary name against the OFAC, EU and UN designation lists. Returns a verdict - stop, review, insufficient_coverage, ok - with every reason attached to its block and its dated source.
Before your agent signs an x402 payment to a counterparty it does not know: submit the 402 challenge you received and get one verdict - ok, review, stop, insufficient_coverage - over the pay-to address screened against the OFAC, EU and UN designation lists, the amount claimed compared with what that resource publishes in the public CDP Bazaar census, and whether the resource is listed at all. Every reason names its block and its source.
Before your agent pays a supplier invoice: send the invoice file itself (Factur-X PDF, CII or UBL XML) and get one verdict - pay, hold, stop, insufficient_coverage - over the document's EN 16931 compliance, the supplier in the INSEE Sirene and BODACC registers, live EU VAT (VIES), sanctions screening of the party actually credited, VAT consistency and the statutory due date. Every reason names the block and the source it comes from.
Run every check on a supplier before you onboard, pay or invoice it, in one call: identity and status (INSEE Sirene or GLEIF), BODACC insolvency and deregistration announcements, live EU VAT validation (VIES), Peppol e-invoicing reachability, the declared control chain, and OFAC/EU/UN sanctions screening of both the company and its ultimate parent. Returns a verdict - stop, review, insufficient_coverage, ok - with every reason attached to its block and source.
Bank-account checks backed by national bank registries, the EPC register and the OFAC/EU/UN & FATF lists — every response carries its provenance.
/iban/resolve
GET /iban/resolve
$0.01
Before your agent pays a new bank account: validate the IBAN (all countries), resolve its bank and BIC from national bank registries, and check SEPA reachability per scheme (SCT, SCT Inst, SDD Core/B2B) against the EPC register
Before a payment run: validate up to 500 IBANs in one call, resolve bank and BIC from national bank registries, and check SEPA reachability per scheme (SCT, SCT Inst, SDD Core/B2B) against the EPC register
Before your agent onboards or pays a counterparty: screen the bank account for sanctions — flags sanctioned banks (BIC) and high-risk jurisdictions against the official OFAC, EU, UN and FATF lists, for KYB/AML compliance
Before onboarding or a payment run: screen up to 500 IBANs in one call for sanctioned banks (BIC) and high-risk jurisdictions against the official OFAC, EU, UN and FATF lists, for KYB/AML compliance
Before your agent pays an IBAN taken from OCR or invoice ingestion: fix a broken or garbled IBAN — recompute check digits, recover unreadable characters (?) and rank candidates by known bank from national bank registries
Before a payment run over OCR or invoice data: fix up to 500 broken or garbled IBANs in one call, within a total repair-effort budget — recompute check digits, recover unreadable characters and rank candidates by known bank from national bank registries. A batch whose combined '?' search space is too large is rejected with 400 BATCH_BUDGET_EXCEEDED and never charged: send fewer IBANs per call, or fewer '?' per IBAN
Sanctions screening backed by the OFAC SDN and Consolidated lists, the EU financial sanctions list and the UN Security Council list — every response carries its provenance.
/screen/entity
GET /screen/entity
$0.01
Before your agent onboards a counterparty, signs a contract or releases a payment: screen a company or person name against the official designation lists in full - OFAC SDN and Consolidated, the EU financial sanctions list, the UN Security Council list - 26,666 entities with their aliases, programmes and designation dates. Deterministic name matching, every hit sourced to the list that published it. PEP screening, adverse media and transaction analysis are not covered.
Before onboarding a supplier book or running a payout list: screen up to 200 company or person names in one call against the official designation lists in full - OFAC SDN and Consolidated, the EU financial sanctions list, the UN Security Council list - one sourced verdict per name, with aliases, programmes and designation dates. A malformed item is flagged at its position instead of failing the batch. PEP screening, adverse media and transaction analysis are not covered.
Before your agent sends funds to a crypto address: check it against the sanctioned addresses published by OFAC (SDN and Consolidated) AND by the EU financial sanctions list - multi-list, where free feeds are OFAC-only - returning the designated entity, its programmes and its designation date behind every hit. Bitcoin, EVM, Tron, Solana and Monero. No transaction analysis, no clustering, no derived addresses, no risk score.
Before a payout run or a sweep: check up to 500 crypto addresses in one call against the sanctioned addresses published by OFAC (SDN and Consolidated) AND by the EU financial sanctions list - multi-list, where free feeds are OFAC-only - each hit carrying the designated entity, its programmes and its designation date. A malformed address is flagged at its position instead of failing the batch. No transaction analysis, no clustering, no risk score.
Phone-number checks backed by libphonenumber (offline) — every response carries its provenance.
/phone/validate
GET /phone/validate
$0.01
Before your agent stores or messages a phone number: validate and normalize it offline — E.164 formatting, region, national format and line type from Google's libphonenumber metadata, no network call
Law-in-force checks backed by LEGI (Légifrance), EUR-Lex and the official public-holiday calendars — every response carries its provenance.
/legal/article
GET /legal/article
$0.01
Before your agent cites, advises on or drafts against a French law: check the article is in force at a given date — consolidated text, status (in force / repealed / deferred) and validity dates, from the official Légifrance LEGI dataset. Covers 12 major French codes and statutes (civil, labor, commercial, tax/CGI and more).
Before your agent advises on EU compliance or cites EU law: check an act or article is in force at a given date, by CELEX or ELI identifier — consolidated text, status, validity dates and the staggered date it becomes APPLICABLE (in force is not the same thing: the AI Act is in force since 2024-08-01, but its Article 50 only applies from 2026-08-02), from the official EUR-Lex dataset. Covers key EU digital and financial acts (GDPR, DSA, AI Act, MiCA and more).
Before your agent signs off a contract review or a due-diligence memo: check that up to 50 legal citations are still in force at a given date in one call — consolidated text, status and validity dates per reference, from the official LEGI (French law) and EUR-Lex (EU law) datasets. Covers 12 major French codes and statutes and key EU acts (GDPR, DSA, AI Act, MiCA and more).
Before your agent relies on a stored legal citation or watches a provision for change: the full version timeline of a French or EU article — each version's status (in force, repealed, superseded, deferred) and validity dates, from the official LEGI and EUR-Lex datasets — to detect exactly when it changed. Covers 12 major French codes and statutes and key EU acts (GDPR, DSA, AI Act, MiCA and more).
Before your agent updates a contract clause or a compliance rule that depends on a law: see exactly what changed in a French or EU article between two dates — resolves the version in force at each date, from the official LEGI and EUR-Lex datasets, and returns a word-level diff plus a git-style unified patch with both versions' metadata. Covers 12 major French codes and statutes and key EU acts (GDPR, DSA, AI Act, MiCA and more).
Before your agent re-certifies a legal knowledge base or monitors a portfolio of citations: the version timeline (status and validity dates) of up to 100 French or EU article references in one call, from the official LEGI and EUR-Lex datasets, to detect which references changed and when. Covers 12 major French codes and statutes and key EU acts (GDPR, DSA, AI Act, MiCA and more).
Before your agent re-runs a contract review or due diligence after time has passed: a structured diff of the consolidated text — word-level segments plus a git-style unified patch — for up to 50 French or EU article references, each between its own two dates, in one call, from the official LEGI and EUR-Lex datasets. Covers 12 major French codes and statutes and key EU acts (GDPR, DSA, AI Act, MiCA and more).
Before your agent commits to a filing date, a notice period or a contractual term: the exact due date of a time limit under an explicitly chosen computation regime — calendar days, calendar days with rollover, business days (Mon-Sat), working days (Mon-Fri), clear days (jours francs), months or years — with every weekend and public holiday it skipped, each one named, the end-of-month and next-working-day rules it applied, and the step-by-step reasoning. Holiday calendars for 10 countries: France from the official State dataset, the nine others from a community reference dataset, and every answer says which authority it used. The regime is required and never inferred from the legal matter.
Before your agent relies on an EU directive producing effects in a Member State: in force, applicable and transposed are three different things. This returns the national implementing measures a Member State has notified to the European Commission for an EU act — each measure with its national act type, title, national reference, publication and notification dates — plus the published transposition deadline and how many days the first notification came after it (positive means late). Ask for one country, or omit it for a one-line verdict on each of the 30 EEA States. What is served is a register of notifications, never an assessment of conformity.
Weather and climate checks backed by Copernicus ERA5, NOAA GFS/GEFS and the WMO 1991-2020 normals — every response carries its provenance.
/climate/point
GET /climate/point
$0.01
Before your agent verifies a weather-dependent claim, backtests a parametric trigger or checks what conditions a site or shipment faced: daily 2 m temperature, total precipitation and 10 m wind at any GPS point or city for a past date, from the Copernicus ERA5 reanalysis — temperature and precipitation cover a rolling 30-year window, wind a rolling ~5-year window — with the served grid cell and a coverage flag.
Before your agent schedules a delivery, plans outdoor work or hedges short-term weather exposure: a point forecast of 2 m temperature, precipitation and 10 m wind — plus opt-in convective, humidity, cloud and snow variables — for any GPS point or city. Deterministic NOAA GFS at 0.25 deg to 16 days (384 h), NOAA GEFS ensemble mean at 0.5 deg to 35 days (840 h) — beyond 16 days the value is an ensemble mean with smoothed extremes: trend, not threshold triggers. The lead selector is optional (omitted, the forecast defaults to the current analysis time) and the response names the model that served it and the exact forecast step.
Before your agent plans a multi-day operation — logistics, construction, an event, crop work: a day-by-day forecast for any GPS point or city, with daily minimum and maximum 2 m temperature, total precipitation and peak 10 m wind per UTC calendar day, from today to the end of the NOAA GFS window (~16 days), extendable to 35 days on the NOAA GEFS ensemble mean — those extra days are an average of ~30 members, so their extremes are smoothed: trend, not threshold triggers. Each day names the model that served it and carries a coverage flag.
Before your agent plans across a whole network — fleet routing, multi-site operations, portfolio hedging: 2 m temperature, precipitation and 10 m wind forecasts for up to 500 point + lead-time queries in one call, from the deterministic NOAA GFS run (~16 days). Each query takes lat+lon or a city name plus an optional lead selector (omitted, it defaults to the current analysis time); an uncovered item is flagged via per-item coverage instead of failing the batch.
Before your agent commits to a rain-or-shine decision two to five weeks out — a parametric payout, a shipping window, a harvest or concrete slot: how many NOAA GEFS ensemble members cross your threshold on each day of the ensemble tail. Ask 'at least 30 mm of rain in a day', 'wind at or above 15 m/s' or 'a day at or below 0 deg C' for any GPS point or city, and get one counted fraction per UTC calendar day with the member count behind it. Ask days= (up to 20) for the whole tail. This is what the ensemble MEAN served by GET /weather/forecast cannot answer: averaging the members smooths extremes away, so a mean day essentially never crosses a threshold. What is served here is a raw counted fraction, not a calibrated probability — no bias correction or statistical post-processing is applied, and members_total is served on every day so the number can be audited.
Before your agent backtests parametric triggers over a site portfolio, verifies claims across parcels or scores weather exposure for an asset fleet: daily 2 m temperature, total precipitation and 10 m wind for up to 500 point/date queries in one call, from the Copernicus ERA5 reanalysis (temperature and precipitation: rolling 30-year window; wind: rolling ~5-year window). Each query takes lat+lon or a city name; an uncovered item is flagged via per-item coverage instead of failing the batch.
Before your agent sizes heating or cooling demand, settles an energy contract or models crop growth: heating, cooling and growing degree-days (HDD/CDD/GDD) for any GPS point or city over a date range, computed from Copernicus ERA5 daily mean temperature (rolling 30-year window) with the NWS/EIA/WMO mean-temperature method.
Before your agent writes a seasonal report, audits an ESG claim or reconciles a weather-linked contract: period climate summaries for any GPS point or city over a date range — mean, min and max temperature, total precipitation and mean wind — from the Copernicus ERA5 reanalysis (temperature and precipitation: rolling 30-year window; wind: rolling ~5-year window), with the served grid cell and a coverage flag.
Before your agent prices or settles a parametric-insurance trigger, or screens weather risk for agriculture and construction: climate-event indices for any GPS point or city over a date range — frost and summer days, tropical nights, dry spells, heavy-rain days, max one-day rainfall, total precipitation and peak gust — ETCCDI/Climdex definitions on true daily Tmin/Tmax from the Copernicus ERA5 reanalysis (precipitation-based indices: rolling 30-year window; Tmin/Tmax- and gust-based indices: rolling ~5-year window).
Before your agent verifies a 'hotter than normal' claim, reports climate exposure or settles an anomaly-based trigger: how far observed temperature, precipitation and wind departed from the WMO 1991-2020 climatological normal, for any GPS point or city over a date range within the ingested Copernicus ERA5 window (rolling 30 years for temperature and precipitation, rolling ~5 years for wind).
Before your agent underwrites, reinsures or stress-tests a climate exposure: rank a target year's season against 30 years of a location's climate history (Copernicus ERA5 reanalysis, WMO-standard climatological period) — percentile, empirical return period and sample size, for any GPS point or city. Temperature and precipitation cover a rolling 30-year window; every verdict reports its actual sample size via sample_years and coverage.
Vehicle checks backed by NHTSA vPIC, RappelConso, the EU CO2 monitoring dataset (Regulation (EU) 2019/631), the French registration-tax schedules and the official Crit'Air tables — every response carries its provenance.
/vehicle/vin/decode
GET /vehicle/vin/decode
$0.01
Before your agent buys, lists or insures a vehicle: decode any 17-character VIN offline — manufacturer, country and region of build, model year, check digit validity and deterministic attributes from the official NHTSA vPIC tables
Before your agent buys, sells or services a vehicle: list manufacturer recall campaigns that may affect it, consolidated from official NHTSA and RappelConso sources, matched by make, model and year or from a VIN. An empty list means no known recall. Results are paged at 200 campaigns, one paid call per page: the busiest brands (~900 campaigns) take 5
Before your agent plans a route or advises on a vehicle in France: compute its Crit'Air class and current low-emission-zone (ZFE) restrictions from the official classification tables (arrêté du 21 juin 2016), from declared fuel and Euro norm or first registration date
Before your agent buys or appraises a vehicle: one consolidated report from a VIN — decode, candidate recall campaigns, type-approval CO2, the French registration tax it would cost and, from declared characteristics, Crit'Air / low-emission-zone status — from official NHTSA vPIC, RappelConso, EU 2019/631 and French tax sources, each block with its own provenance and honest coverage
Before your agent prices, imports or compares a European vehicle: get its type-approval CO2 emissions and technical characteristics — WLTP g/km with its dispersion, fuel, engine capacity, power and mass — from the official EU CO2 monitoring data (Regulation (EU) 2019/631, new registrations 2018-2022), keyed by type-approval number or by make, model and year. Every answer states the precision it was matched at
Before your agent buys, imports or quotes a vehicle in France: compute what registering it costs — CO2 malus, weight malus and the regional certificate tax — against the official schedule in force on the registration date you supply, not the one a model remembers. Coverage is France only; each item states its bracket, its threshold and the exemption applied with its legal article, and the total disappears rather than being partial
Before your agent buys a used vehicle: one VIN returns a decision - ok, review, stop, insufficient_coverage - over open recall campaigns (NHTSA, RappelConso), type-approval CO2 from the European Environment Agency, Crit'Air low-emission-zone class and what the French registration certificate would cost. Every reason names its block; the answer states what it cannot attest - service history, mileage, liens, accident damage.
Turn a vehicle designation an agent actually holds - 'Renault Clio V', 'VW Golf 2020' - into the make/model/year key that /vehicle/recalls and /vehicle/co2 require, spelled the way the referential spells it. Sources: NHTSA and RappelConso recall campaigns, EEA European registration aggregates. Closed verdict: exact, ambiguous or none. Several model years for one model is the normal answer, and every candidate is returned - we never pick one for you.
E-invoice checks backed by the EN 16931 semantic model and the official Factur-X, CII and UBL specifications, plus the fiscal verdicts of Directive 2006/112/EC (VAT) and Directive 2011/7/EU (late payment) — every response carries its provenance.
/invoice/read
POST /invoice/read
$0.01
Read any EU structured e-invoice in one call: post a Factur-X PDF, CII or UBL XML and get faithful structured data - parties with SIREN/VAT identifiers, line items, VAT breakdown, totals, payment IBAN and due dates. Covers the three formats mandated by the French 2026-2027 e-invoicing reform (EN 16931). No account, no key, no OCR guesswork.
Should you accept this invoice? Post a Factur-X PDF, CII or UBL XML and get a deterministic EN 16931 verdict: XSD conformance, every violated business rule by its official BR id with observed vs expected values, Factur-X profile consistency, and syntax checks on IBAN, SIREN/SIRET and VAT identifiers. The formats mandated by the French 2026-2027 reform. A non-conformant invoice is a paid, documented answer - no LLM guesswork. Add ?ruleset=xrechnung or ?ruleset=peppol to layer a national CIUS pack on top of the EN 16931 base: XRechnung (German BR-DE rules, DE mandate) or Peppol BIS Billing 3.0 (BE mandate, live since 2026-04), each violation sourced by its official id (BR-DE-xx, PEPPOL-EN16931-Rxxx). Each verdict also carries informational VAT consistency checks (INV-VAT-*), never mixed with the normative BR ids.
Before your agent bills a customer: get the invoice document itself, ready to send and archive - not a verdict on someone else's file. Post simple business JSON (parties, lines, VAT), receive XML already cleared by the full official EN 16931 rule set, totals computed, inconsistent input refused field by field. CII by default (Factur-X EN 16931 URN), syntax=ubl for UBL 2.1, ruleset=peppol for Peppol BIS Billing 3.0. vat_treatment writes the reverse charge or intra-Community exemption mention.
Before your agent bills a customer: get the whole Factur-X file it sends and the customer reads - a PDF/A-3B embedding the EN 16931 CII XML, the hybrid invoice mandated by the French 2026-2027 reform. Post simple business JSON (parties, lines, VAT), get a file ready to send and archive: totals computed, inconsistent input refused field by field, the official rule set cleared and the embedded XML re-read byte for byte. vat_treatment writes the reverse charge or intra-Community exemption mention.
Before your agent issues or books a cross-border invoice: who charges the VAT, under which mechanism, where the supply is taxed and with which compulsory wording. Post one transaction (countries, goods or services, business or consumer customer, supply kind, date, optional amount) and get a determination sourced article by article under Directive 2006/112/EC: place of supply, person liable, standard VAT, reverse charge, intra-Community exemption or out of scope, the rate in force at that date, and the exact invoice mentions to print. An invoice missing its reverse-charge mention passes the EN 16931 rules and is still refused by the tax authority - the check /invoice/validate cannot make. The customer's VAT number is taken AS DECLARED (zero network): verify it live with /company/vat. Uncovered regimes (OSS/IOSS, import VAT, margin schemes, triangulars) come back as a motivated undetermined, never a plausible default.
A sales ledger or an order book judged in one call: up to 200 transactions, one VAT determination each, one settlement. Same verdicts as /invoice/vat-treatment - place of supply, person liable, standard VAT, reverse charge, intra-Community exemption or out of scope, the rate in force at the date of each line, the compulsory invoice mentions and the articles of Directive 2006/112/EC they rest on. A malformed item is reported at its position instead of failing the batch; only an empty or oversized batch is a 400. Legal bases and caveats are shared across the batch and cited by key, and the summary counts the five mechanisms. Customer VAT numbers are taken AS DECLARED (zero network by design) - verify them with /company/vat; uncovered regimes (OSS/IOSS, import VAT, margin schemes, triangular transactions) come back as a motivated undetermined, never a plausible default.
Before your agent sends a dunning letter or books a receivable: what is legally due on an unpaid B2B invoice. Post the claim (amount, currency, country, due date - or invoice date plus agreed terms - and payment date) and get late-payment interest computed PERIOD BY PERIOD under Directive 2011/7/EU: the reference rate changes every 1 January and 1 July, so a claim straddling several half-years is split into tranches, each at the rate of its own half-year - applying today's rate to the whole period is simply wrong. Statutory rate = ECB main refinancing rate (NBP reference rate for Poland) plus the national margin, with the national transposition actually applied: France 10 points, Germany 9 on the Basiszinssatz base, Belgium rounding the reference up to the half point, 8 points elsewhere. Covers FR, DE, ES, IT, NL, BE, PT and PL, with an explicit fallback to the Directive minimum for the other Member States - a national rule is never guessed. Adds the fixed recovery compensation of Art. 6 (EUR 40, tiered in Poland), each component cited to its article. Paying on time is a valid answer, not an error.
Counterparty due-diligence checks backed by the official INSEE Sirene register, the BODACC gazette, VIES, the Peppol directory and the global GLEIF LEI reference — every response carries its provenance.
/company/resolve
GET /company/resolve
$0.01
Verify a French company before you invoice, pay or onboard it: legal name, active/ceased status, legal form, NAF activity, workforce band, creation date and head-office address from the official INSEE Sirene register (30M+ legal units, monthly snapshot). Lookup by SIREN or SIRET. No account, no key, pay per call.
Check a French company for official insolvency, deregistration and business-sale announcements before you invoice or extend credit: dated event timeline and factual flags from the official BODACC gazette, by SIREN. An empty list is a paid answer: no known announcement over the covered window. No account, no key.
Validate any EU VAT number live against the official European Commission VIES service before you invoice cross-border, or derive and check the French VAT number from a SIREN. Returns valid/invalid or an honest 'unverifiable' when the member state is down - syntax check always included. No account, no key.
Check whether a company can receive e-invoices on the Peppol network before you send one: participant identifiers and accepted document types from the official Peppol Directory export, by SIREN, SIRET or explicit Peppol participant ID. 'Not registered' is a paid, dated answer. No account, no key.
Full counterparty due-diligence on a French company in one paid call, by SIREN: Sirene identity and status, dated BODACC insolvency/deregistration/sale announcements, live EU VAT validation (VIES) and Peppol e-invoicing reachability - each block with its own provenance. The pre-action check before you invoice, pay or onboard. No account, no key.
Identify the company behind a bank identifier, a French SIREN, an ISIN or an exact name before you onboard, pay or invoice it: official GLEIF LEI identity - legal name, jurisdiction, entity and registration status, addresses, national registration number, mapped BICs and ISINs. Lookup by LEI, BIC (8 or 11), ISIN, SIREN, or exact legal name plus country. Homonyms are returned as candidates, never silently resolved.
Find out who controls a counterparty before you onboard, pay or invoice it: the declared control chain from the official GLEIF Level 2 reference, by LEI - direct parent, ultimate parent, the dated chain between them, and direct subsidiaries. A declared absence of parent comes back with its official reason, never as an empty block. This is the accounting consolidation parent, not beneficial ownership.
Medication checks backed by the French BDPM (ANSM/HAS) register and the US openFDA and RxNorm datasets - every response carries its provenance.
/medication/resolve
GET /medication/resolve
$0.01
Resolve a medication before your agent acts on a prescription or an order: official identity from the French BDPM registry (ANSM/HAS) by CIS code, exact name or DCI substance - authorization and marketing status, active substances normalized to INN (the FR-US bridge), pack-level price and reimbursement rate, generic group. Same endpoint resolves US drugs by NDC or name (openFDA/RxNorm). No account, no key, pay per call.
Find the French BDPM generic group of a medication before your agent substitutes or sources it: given a CIS code, exact name or DCI substance, returns the group's principle brand (princeps) and every equivalent generic, each with its role (princeps/generique), official name and pack-level public price and reimbursement rate when the BDPM publishes them. French (BDPM) scope: a US NDC key returns an honest FR-only coverage note. No account, no key, pay per call.
Check a medication's availability before your agent orders or substitutes it: shortage reports from the French ANSM (data.ANSM) by CIS code or substance (INN/DCI) - status (rupture, risque_rupture, tension, remise_a_disposition, arret_commercialisation, or no_known_shortage), report dates, expected return and the official ANSM report URL. Honest by design: ANSM only records reported shortages, so the absence of a report is never a stock guarantee. French market scope: a US NDC key returns an honest FR-only coverage note. No account, no key, pay per call.
Check drug-drug interactions against the authoritative French ANSM thesaurus before your agent acts on a prescription or an order: post 2+ medications (CIS code, US NDC, INN/DCI substance or ANSM class) and get every published interaction between them - constraint level (contre_indication, association_deconseillee, precaution_emploi, a_prendre_en_compte), management advice and mechanism exactly as published by ANSM, plus honest coverage of anything that could not be compared. US drugs join via the INN substance bridge. No account, no key, pay per call.
One consolidated medication report before your agent acts on a prescription or an order, by French CIS code or US NDC: official BDPM identity with pack-level price and reimbursement, INN-normalized substances, BDPM generic group, ANSM shortage status, and ANSM-thesaurus drug-drug interactions between the subject and an optional co-medication list (with=, comma-separated CIS/NDC/DCI). Every section carries its own source and date; honest coverage. No account, no key, pay per call.
Turn a medication name an agent actually holds - 'Doliprane 1000', 'Tylenol', 'paracetamol' - into the CIS code, product NDC or canonical active-substance identifier that /medication/resolve, /medication/interactions, /medication/generics and /medication/availability require. Sources: the French BDPM register and openFDA/RxNorm. Closed verdict: exact, ambiguous or none. One brand name covering several strengths and forms is the normal answer, and every candidate is returned - we never pick one for you.
EU labelling verdicts backed by the ECHA Annex VI harmonised classification (CLP) versioned by ATP, the cosmetics Regulation (EC) No 1223/2009 and the food-additive Union list of Regulation (EC) No 1333/2008 — every response carries its provenance.
/label/chem/classify
GET /label/chem/classify
$0.01
Before your agent approves a substance for an EU label or safety data sheet: the harmonised classification and labelling from Annex VI of the CLP regulation (ECHA), by CAS, EC number or exact name - hazard classes with H statements, GHS pictograms, signal word, concentration limits, versioned by ATP: the entry in force at the queried date, plus the upcoming change when a published but not-yet-applicable ATP modifies it. Absence from Annex VI returns an explicit self-classification verdict.
Before your agent clears a BOM or a formulation for the EU market: screen up to 200 substances in one call against Annex VI of the CLP regulation (ECHA) - one harmonised-classification verdict per item (CAS, EC number or exact name) with H statements, GHS pictograms, signal word, and the ATP version in force plus upcoming ATP changes; a malformed item is flagged at its position instead of failing the batch, and an unlisted substance gets an explicit self-classification verdict.
Before your agent clears a substance, an article or a supply-chain declaration for the EU market: is it on the ECHA Candidate List of substances of very high concern (SVHC, REACH)? By CAS, EC number or exact name - the published entry with its Article 57 inclusion reasons (CMR, PBT, vPvB, equivalent concern), inclusion date and the ECHA inclusion decisions as published, plus the generic obligations it triggers (REACH Article 33 information duty, SCIP notification above 0.1 % w/w). A not-listed verdict is dated to the served list, never an all-clear.
Before your agent clears a BOM, a formulation or a supplier declaration for the EU market: screen up to 200 substances in one call against the ECHA Candidate List of SVHC (REACH) - one verdict per item (CAS, EC number or exact name) with the published Article 57 inclusion reasons, inclusion date and the triggered obligations (REACH Article 33, SCIP notification); a malformed item is flagged at its position instead of failing the batch, and a not-listed substance gets an explicit verdict dated to the served list.
Before your agent puts a nutrition or health claim on an EU food label: is the proposed wording an authorised claim (exact legal wording + conditions of use), a non-authorised one (official rejection reason + EFSA opinion), or absent from the EU Register (Regulation (EC) 1924/2006)? Post the label wording (and optionally the nutrient); matching is strict-normalised (case, punctuation, spacing), exact or containment, never fuzzy - zero false positives. The Register carries the REJECTED claims a plain LLM validates by mistake. An absent wording returns the authorised claims for the nutrient plus a limit note (absence is neither authorisation nor prohibition - 'on hold' botanicals are not in the Register).
List the authorised EU nutrition and health claims for a given nutrient or substance (Regulation (EC) 1924/2006): what your agent may legally write on a food label. Pass the nutrient; get every authorised claim as published - exact legal wording, conditions of use, EFSA opinion and legislation. A known nutrient whose claims are all rejected returns an explicit empty list with the rejected count (not a silent gap); a nutrient absent from the Register returns an empty list with a limit note (absence is neither authorisation nor prohibition - 'on hold' botanical claims are not in the Register). The counterpart of /label/claims/check: check starts from a wording, permitted starts from the nutrient.
Before your agent clears an ingredient for an EU cosmetic label: is it prohibited (Annex II), restricted (Annex III), or authorised as a colorant (IV), preservative (V) or UV filter (VI) under Regulation (EC) No 1223/2009? By CAS, EC number, INCI name or exact chemical name - every published entry with its maximum concentration, conditions of use and the MANDATORY LABEL WORDING served word for word, versioned by amending regulation: the consolidated version in force at the queried date plus the upcoming change when a published but not-yet-applicable one modifies the entry. CosIng is a legally non-binding information tool: only the Annexes are authoritative, and any divergence is reported, never silently resolved. Absence from the Annexes is not an authorisation - the verdict says so explicitly.
Before your agent clears a cosmetic formulation for the EU market: screen a whole INCI list - up to 200 ingredients in one call - against Annexes II to VI of Regulation (EC) No 1223/2009. One verdict per item (prohibited, restricted, authorised colorant/preservative/UV filter, or not listed) with the published entries, maximum concentrations, conditions and MANDATORY label wording served word for word, plus a summary count per verdict. A malformed item or an unresolved identity is flagged at its position instead of failing the batch; the input order is preserved but no conclusion is drawn from it (an INCI list's descending-concentration order is not exploited). This is not a formulation compliance verdict: no concentration in the finished product is computed.
Before your agent validates a recipe, a food label or an import: is this additive authorised in this food category, at what maximum level and under which conditions (Union list, Regulation (EC) No 1333/2008, Annexes II and III)? By E number or exact published name, WITH a food category code - you get an authorised/not_authorised verdict with the maximum level, restrictions and footnotes served word for word ('quantum satis' as published, never converted), and via_group when the authorisation comes from a published group such as 'E 249 - 250'. WITHOUT a food category you get the inventory of categories where it is authorised and NO verdict: an additive has no status in the abstract. A not_authorised verdict is dated - the Union list is positive. Where the text bounds a line by a 'Period of application', the lines that apply at your date come first and each one carries in_force_at_date: an expired line is flagged, never dropped.
Turn an ingredient name read off a label - 'sodium nitrite', 'titanium dioxide', an INCI name - into the exact CAS, EC, INCI or E number that /label/chem/classify, /label/chem/svhc, /label/cosmetic/ingredient and /label/food/additive require. Sources: ECHA Annex VI and Candidate List, the consolidated Cosmetics and Food Additives Regulations, and the Commission CosIng index. Closed verdict: exact, ambiguous or none, and every candidate names the endpoints it opens.
Additional checks from the live catalog — every response carries its provenance.
/certify/batch
POST /certify/batch
$0.05 + $0.02/subject
Clean a customer or supplier file before the French e-invoicing switchover: submit up to 200 lines with their SIREN or SIRET and get one sourced verdict per line - existence and administrative state in the INSEE Sirene register, insolvency and deregistration announcements in the BODACC gazette, and the name you hold compared to the one the register publishes. The batch carries a signed Merkle root, so the buyer of the file can verify any single line offline.
Screen a file of counterparties against sanctions lists and keep the proof: up to 200 names or crypto addresses in one call, each compared to the OFAC, EU and UN lists with the publication date of every list served. The batch carries a signed Merkle root, so whoever receives your file verifies a single line offline - what was screened, against which lists, on which dates.
