Skip to main content

agentdata

PyPI License Python Commands

One Python client and CLI over 256 public datasets — company registries, sanctions lists, procurement awards, drug approvals — with no API keys and no accounts.

Government data holds the decisive facts about companies, but it is scattered across 173 public APIs, each with its own pagination, rate limits, and undocumented quirks. agentdata puts 256 of those datasets behind one interface: a pure-standard-library client and CLI across 35 jurisdictions. Every upstream gotcha we hit live is already pinned — politeness and caching are built in.

Install

python -m pip install agentdata

For a reproducible install, pin the release: python -m pip install agentdata==0.3.68.

Prefer not to touch PyPI? Install straight from a release tarball:

python -m pip install https://github.com/dicaria/agentdata/archive/refs/tags/v0.3.68.tar.gz

Or via git (pin a tag for reproducible installs):

python -m pip install git+https://github.com/dicaria/agentdata.git@v0.3.68

Or track the latest release (convenience for agents and scripts — a fresh release does not trigger a reinstall under the same URL, so pass --force-reinstall when you want the new one):

python -m pip install --force-reinstall https://github.com/dicaria/agentdata/releases/latest/download/agentdata-latest.tar.gz

Two API families

Family Where it runs Credentials Cost
Enrichment Locally — the SDK calls public government / open-data APIs from your machine none free
Corpus Hosted service operated by dicaria credentials issued by dicaria per service agreement

Most commands are enrichment: they need no configuration, no keys, no account. Corpus commands are marked in the docs and require credentials issued by dicaria under a separate service agreement — contact contact@dicaria.com.

What can you ask it?

Commands follow <domain> <source>_<dataset> — one command per upstream dataset, parameters kept under their upstream names. Every row below works as agentdata <domain> <command> … from the shell, or c.<domain>.<command>(…) from Python.

Question Commands
Is this company sanctioned? legal csl_screening_list · legal co_sic_sanctions · legal wb_debarred_firms · legal ca_sema_sanctions · legal ru_cbr_warning_list
Do consumers complain about this company? legal cfpb_complaints · legal cfpb_suggest_company · legal br_bacen_complaints
Who owns or controls it? company gleif_direct_parent · company no_brreg_roller · company fr_recherche_search
Did it win public contracts? procurement br_pncp_contracts · procurement ungm_contract_awards · procurement wb_procnotices
Is its drug or device approved — or recalled? healthcare fda_drug_approvals · healthcare gb_mhra_alerts · healthcare fda_drug_recalls
What does this facility or fleet emit? esg climatetrace_sources · transportation thetis_mrv_datasets · transportation paris_mou_detentions
Which aircraft does this registration mark belong to? transportation ca_ccar_search · transportation ca_ccar_aircraft
Pull a company card from a national registry company ch_zefix_firm · company co_rues_companies · company kr_hometax_biz_status · company kz_gbd_ul_companies · company cz_ares_ekonomicke_subjekty · company in_sebi_intermediaries · company tr_spk_listed · company tr_bddk_institutions · company ru_cbr_credit_orgs · company br_bacen_institutions · company br_bacen_branches · company br_bacen_domains · company br_bacen_dasfn
Has its broker been expelled, or its regulator fined it? legal in_sebi_orders · company in_sebi_default_brokers
Has the regulator fined it, banned it, or put it under a special regime? legal tr_spk_fines · legal tr_spk_bans · legal br_bacen_regimes
How big is the country it operates in? economy wb_wdi_values · economy eurostat_table · economy fred_series
Convert an EUR amount, or read a bilateral trade flow economy ecb_fx_rates · economy un_comtrade_trade

Quick start

from agentdata import Client

c = Client()  # zero config: enrichment commands work immediately

c.healthcare.fda_drug_recalls(search='recalling_firm.exact:"SV Labs Toronto Corporation"')
c.company.gleif_lei_records(lei="5493001KJTIIGC8Y1R12")           # LEI -> entity identity card (GLEIF)
c.company.ch_zefix_firm_search(name="siemens")                    # Swiss commercial registry (Zefix)
c.esg.climatetrace_sources(owner_ids="E100000001492", year=2023)  # owner -> facility-level emissions
c.transportation.paris_mou_inspections(period_min="01/12/2025", period_max="31/12/2025", ism_company_imo_number="0052193")  # port-state inspection flow by management company (Paris MoU; dd/mm/yyyy window up to 3 years)
c.legal.csl_screening_list(name="VTB BANK")                       # sanctions screening (US CSL)
c.research.openalex_works(search="vaccine", ror="01xdqrp08")
c.procurement.br_pncp_contracts(data_inicial="20260901", data_final="20260930", pagina=1)
c.security.osv_vulns(package="lodash", ecosystem="npm", version="4.17.15")
c.economy.wb_wdi_values(country="CHN", indicator="NY.GDP.MKTP.CD", date="2020:2023")  # World Bank indicator values
c.economy.ecb_fx_rates(currency="USD", start="2026-10-01", end="2026-10-31")  # euro reference rate (1 EUR = X)

The same commands from the shell — ten entry points, agentdata plus one per domain:

agentdata healthcare rxnav_rxcui_by_name --name atorvastatin
agentdata company gleif_lei_records_search --filter-entity-legal-name "Siemens" --page-size 3
agentdata legal csl_screening_list --name "VTB BANK"
agentdata esg climatetrace_sources --owner-ids E100000001492 --year 2023
agentdata procurement br_pncp_contracts --data-inicial 20260901 --data-final 20260930 --pagina 1
agentdata economy wb_wdi_values --country CHN --indicator NY.GDP.MKTP.CD --date 2020:2023

If a domain entry-point command (company, legal, …) collides with another tool on your machine, python -m agentdata <domain> … is the equivalent invocation.

Corpus commands use the same shape with hosted credentials:

c2 = Client(base_url="https://…", api_key="mk_live_…")  # PAT issued by dicaria
c2.company.read(id=1476, level="tier2")

Show me a response

Real output of agentdata healthcare rxnav_rxcui_by_name --name atorvastatin — every command returns this shape:

{
  "ok": true,
  "domain": "healthcare",
  "tool": "rxnav_rxcui_by_name",
  "input": { "name": "atorvastatin" },
  "from_cache": false,
  "reason": "RxNav normalization: drug name 'atorvastatin' → RxCUI 83367 (ingredient level)",
  "result": {
    "rxcui": "83367",
    "normalized_name": "atorvastatin",
    "tty": "IN"
  },
  "evidence": [
    {
      "quote": "RxCUI 83367 = atorvastatin",
      "url": "https://rxnav.nlm.nih.gov/REST/rxcui/83367/properties.json"
    }
  ],
  "source": {
    "provider": "rxnav",
    "execution": "local",
    "observed_at": "2026-11-02T03:28:22.213Z"
  },
  "request_id": "3f343e2f-3b08-46e8-ac69-8ee05adce548",
  "jurisdiction": "US"
}

A plain-language reason, evidence rows with a checkable quote and source URL, provenance in source, the issuing authority in jurisdiction, and a request_id for support — your code parses one shape across all datasets. Cached responses are identical, with from_cache: true.

Why not just call the APIs yourself?

You can — and then you get to rediscover what is already pinned here:

  • Quirks are pre-solved, not rediscovered. Every command encodes traps we hit against the live API: anti-forgery token dances, Socrata parameters that are silently ignored, registries that clamp your page size without an error.
  • One envelope across all 256 commands. Query echo, reason, evidence, source metadata, jurisdiction — your code parses one shape.
  • Upstream parameter names, verbatim. What the official API docs call a parameter is what the SDK takes — no invented vocabulary to unlearn.
  • Polite by default. Per-source rate limits are enforced before every request (rate-limits ledger), and a cache hit costs no rate budget.
  • Zero dependencies. Python 3.10+ standard library only.

Coverage

256 datasets · 173 upstream providers · 35 jurisdictions: US 64 · global 40 · EU 18 · UK 15 · Canada 12 · Brazil 10 · France 10 · JP 9 · ES 6 · India 6 · Türkiye 6 · Switzerland 5 · IT 5 · Korea 5 · Australia 4 · DE 4 · Colombia 3 · MX 3 · NL 3 · AT 2 · Czechia 2 · EE 2 · Guernsey 2 · Ireland 2 · LV 2 · Norway 2 · New Zealand 2 · PL 2 · RU 2 · SG 2 · UA 2 · DK 1 · Finland 1 · Jersey 1 · Kazakhstan 1.

Domain Commands Sources
company 79 GLEIF, SEC EDGAR (registry, full-text search & the XBRL structured-facts family), FDIC, USAspending, TED, VIES, UK Contracts Finder & FTS, Zefix (CH), Brreg (NO), YTJ (FI), recherche-entreprises (FR), RUES (CO), Hometax (KR), ROR, NZ Companies Office, ARES (CZ), GFSC (GG), JFSC (JE), GBD_UL (KZ), SEBI intermediaries (IN), SPK public-company register (TR), BDDK licensed institutions (TR), CBR credit institutions (RU), BACEN institutions, branches, verified domains & open-finance API directory (BR), federal insolvency announcements (DE), FSA licensed-institution registers (JP), CCAA restructuring roster (CA), CNMV supervised entities & BORME gazette (ES), data.gov.sg platform (SG), TDLR licenses (TX) & DOH credentials (WA), Chicago BACP business licenses, LA City business register, NYC DCWP licenses, ACPR Regafi financial-sector register (FR), J-PlatPat IP trio & KIPRIS patents (JP/KR), Unternehmensregister register search (DE), KAP members & disclosures (TR), OEPM BOPI gazette index (ES), INPI accounting ratios (FR), EIOPA IORP pension register (EU), BNDES financing operations (BR)
economy 18 World Bank WDI, Eurostat, ECB reference rates, FRED, WHO xMart, UN Comtrade, ILOSTAT, BLS, BIS, IMF, OECD, Comext
esg 8 Climate TRACE, EITI, EU ETS (EUTL), EPA FRS, UK FSA food-hygiene ratings (FHRS), SGM mining concessions (MX)
healthcare 34 US FDA approvals & recalls, ClinicalTrials.gov, RxNav, DailyMed, AccessGUDID, UK MHRA alerts, Health Canada DPD & MDALL, EUDAMED, RASFF & Safety Gate (EU), TGA (AU), Europe PMC, MeSH, US NPPES provider organizations, NAVII facility search (JP), CNES facility register (BR)
legal 87 US consolidated screening list, eCFR, Federal Register, LDA filings, CourtListener, DG COMP, UK CMA cases, UK Electoral Commission, The Gazette, Irish Revenue defaulters, US OCC enforcement, TNA case law, SIPRI, SIC sanctions (CO), MDB debarment rosters (WB/AfDB/IDB), Canada SEMA, CFPB consumer complaints, SEBI enforcement orders (IN), SPK administrative fines & trading bans (TR), CBR illegal-finance warning register (RU), BACEN special regimes & complaint rankings (BR), RIS case law & consolidated federal law (AT), FERC eLibrary (US), French competition-authority fines (FR), Italian ARERA decisions (IT), CFTC enforcement actions (US), Irish lobbying register (IE), Estonian party-finance disclosures (EE), Ukrainian party-asset register (UA), CNBV sanctions & disciplined servants (MX), Rechtspraak case law (NL), KR statutes & court precedents, BOE gazette table of contents (ES), ASC enforcement & BCSC disciplined (CA), AMF sanctions & transactions (FR), CJEU case-law search (EU), KNAB party finance (LV), EFK party financings & FINMA enforcement news (CH), India Code (IN), Federal Register of Legislation (AU), CADE competition cases (BR), CBSA advance rulings (CA), Bundestag lobby register (DE), Danish FSA supervision archive, AGCM proceedings trio (IT), UOKiK decisions (PL), CBP CROSS rulings (US), administrative-courts judgment archives (FR), OCC EASearch enforcement database (US), ACM publications (NL), CNMV sanctions (ES), FSS sanctions (KR), JFTC decision database (JP), CCI antitrust orders & combinations registers (IN), ACCC actions feed (AU), NMW naming-scheme rounds & single-round dossiers (UK), CFTC RED List & reparation decisions (US), CFTC OGC opinions & adjudicatory orders (US)
procurement 7 PNCP (BR), SECOP II (CO), UNGM, World Bank, New York Authorities / MTA / OGS (US)
research 10 OpenAlex, NIH RePORTER, NSF, Crossref, UKRI GtR, iCite, bioRxiv, OpenCitations
security 4 OSV, CISA KEV, ENISA EUVD
transportation 9 Canada TC civil aircraft register (CCAR), EU Air Safety List, Paris MoU port-state control, THETIS-MRV (EU shipping), US FMCSA motor-carrier census

Full command index

Ten entry points — agentdata plus one command per domain — and 256 dataset commands under them: one command per upstream dataset, parameters kept under their upstream names. The block below is the complete index, grouped by domain in DOMAIN_ORDER — the order the CLI registers the domains in (the 0.3.10 ordering ruling). Every line is a runnable command.

Above them sit two meta-commands that are not dataset faces: agentdata capabilities (discovery, since 0.3.52 — renders the full face table from the command tree: --format json|md, --domain-filter, --face; no key, works offline) and the agentdata hub family (PAT → short-lived subject JWT, see above).

agentdata capabilities --format md

# company (79)
agentdata company gleif_lei_records --lei 5493001KJTIIGC8Y1R12
agentdata company gleif_lei_records_search --filter-entity-legal-name Nestle --filter-entity-jurisdiction CH --page-size 5
agentdata company gleif_direct_parent --lei 5493001KJTIIGC8Y1R12
agentdata company gleif_ultimate_parent --lei 5493001KJTIIGC8Y1R12
agentdata company eu_eiopa_iorps --limit 5
agentdata company fdic_failures --filters PSTALP:TX --limit 5
agentdata company fdic_financials --filters CERT:628 --sort-by REPDTE --sort-order desc
agentdata company fdic_history --filters CERT:628 --limit 3
agentdata company fdic_institutions --filters CERT:628
agentdata company fdic_locations --filters STALP:TX --limit 5
agentdata company us_sec_submissions --cik 320193
agentdata company us_sec_companyconcept --cik 320193 --concept RevenueFromContractWithCustomerExcludingAssessedTax
agentdata company us_sec_companyfacts --cik 320193
agentdata company us_sec_frames --concept NetIncomeLoss --frame CY2024
agentdata company us_sec_tickers --ticker AAPL
agentdata company ror_organizations --query EMBL
agentdata company vies_vat --vat DE811907980
agentdata company usaspending_spending_by_award --recipient-search-text "Lockheed Martin" --start-date 2024-01-01 --end-date 2024-12-31 --limit 3
agentdata company ted_notices_search --query 'organisation-name-buyer=Siemens* AND organisation-country-buyer=DEU' --limit 5
agentdata company gb_fts_release_packages --name nhs --updated-from 2026-10-01 --updated-to 2026-10-27 --limit 5
agentdata company gb_contracts_finder_published_notices --name council --published-from 2026-10-20 --published-to 2026-10-27 --limit 5
agentdata company us_sec_fulltext_search --q "artificial intelligence" --forms 8-K --limit 5
agentdata company ch_zefix_firm_search --name "Nestle" --max-entries 3
agentdata company ch_zefix_firm --ehraid 347554
agentdata company no_brreg_enheter --navn equinor --size 2
agentdata company no_brreg_roller --organisasjonsnummer 923609016
agentdata company nzco_companies --q "air new zealand" --limit 15
agentdata company nzco_directors --company-number 104799
agentdata company fi_ytj_companies --name nokia
agentdata company fr_inpi_ratios --siren 844516393 --limit 5
agentdata company fr_inpi_ratios_sectors --classe-naf 1102 --exercice 2019
agentdata company fr_recherche_search --q danone
agentdata company fr_recherche_search --q "552032534" --per-page 25 --etat-administratif A
agentdata company fr_regafi_entities --q Lille --limit 3
agentdata company fr_regafi_entities --siren 265907741
agentdata company fr_regafi_authorisations --id-referentiel 79 --limit 3
agentdata company co_rues_companies --where "razon_social like '%BANCOL%'" --limit 5
agentdata company kr_hometax_biz_status --biz-no 124-81-00998
agentdata company kr_kipris_patents --q 삼성전자 --page-size 3
agentdata company kz_gbd_ul_companies --text "ВЕЛЛЕТ" --count 5
agentdata company cz_ares_ekonomicke_subjekty --ico 45274649
agentdata company cz_ares_ekonomicke_subjekty_vyhledat --obchodni-jmeno "ČEZ" --pocet 5
agentdata company gg_gfsc_registered_businesses --category accountant --limit 5
agentdata company gg_gfsc_regulated_entities --licence "Investment Licensee" --status active --limit 5
agentdata company je_jfsc_regulated_entities --keyword Barclays --page-size 5
agentdata company in_sebi_intermediaries --category sb_equity --search MOTILAL
agentdata company in_sebi_default_brokers
agentdata company tr_spk_listed --segment borsadisi --name AKSEL
agentdata company tr_spk_listed --mersis 0470003286100012
agentdata company ru_cbr_credit_orgs --name "ДЕМО" --status active --limit 5
agentdata company ru_cbr_credit_orgs --ogrn 1037700000001
agentdata company br_bacen_institutions --family banco --segmento múltiplo --limit 5
agentdata company br_bacen_branches --if-name CAIXA --uf SP --limit 5
agentdata company br_bacen_domains --tipo none --limit 5
agentdata company br_bacen_dasfn --api pix_saque --limit 5
agentdata company br_bndes_financiamento --name PETROBRAS --rows 5
agentdata company tr_bddk_institutions --category bankalar --limit 5
agentdata company tr_bddk_institutions --status exited --limit 5
agentdata company tr_kap_members --type IGS --state A
agentdata company tr_kap_disclosures --from 2026-11-01 --to 2026-11-10 --limit 5
agentdata company de_inso_search --state NW --date-from 2026-11-02 --date-to 2026-11-02 --limit 5
agentdata company de_inso_register_lookup --register-court "Düsseldorf" --register-type HRB --register-number 101036 --date-from 2026-11-02 --date-to 2026-11-02 --limit 5
agentdata company de_register_search --q "Siemens Energy AG"
agentdata company jp_fsa_licensed --category crypto_exchanges --limit 5
agentdata company jp_fsa_licensed --category trust_agents --name 北洋 --limit 5
agentdata company jp_design_search --q ソニー
agentdata company jp_patent_search --q ソニー --year 2024 --year-sign ">="
agentdata company jp_trademark_search --q ソニー
agentdata company ca_ccaa_records --province QC --limit 5
agentdata company ca_ccaa_records --name "MTE Logistix"
agentdata company ca_ccaa_records --monitor KPMG --province AB --limit 5
agentdata company es_borme_day --date 2026-11-04 --limit 5
agentdata company es_borme_day --date 2026-11-04 --kind A --limit 3
agentdata company es_borme_notice --id BORME-A-2026-181 --order 2
agentdata company es_borme_notice --id BORME-C-2026-5782
agentdata company es_cnmv_entities --category securities_firms --limit 5
agentdata company es_cnmv_entities --category securities_firms --nif A-83217281
agentdata company es_cnmv_entities --category eea_investment_firms_passport --name Goldman --limit 3
agentdata company es_oepm_bopi --tomo 1 --date 2026-11-04
agentdata company sg_datagovsg_datasets --limit 5
agentdata company sg_datagovsg_datasets --q ACRA --limit 5
agentdata company sg_datagovsg_dataset_download --id d_3f960c10fed6145404ca7b821f263b87
agentdata company sg_datagovsg_dataset_download --id d_b1d2b840ab9e993570c037b706b39bb8
agentdata company tx_tdlr_licenses --limit 3
agentdata company tx_tdlr_licenses --where "license_type='Cosmetology Operator' and business_county='TRAVIS'" --limit 3
agentdata company wa_doh_credentials --limit 3
agentdata company wa_doh_credentials --where "credentialtype='Physician And Surgeon License' and status='Active'" --limit 3
agentdata company chi_bacp_licenses --where "license_description='Consumption on Premises - Incidental Activity' and license_status='AAI'" --limit 3
agentdata company lacity_registrations --where "naics like '7213%' and council_district='10'" --limit 3
agentdata company nyc_dcwp_licenses --where "business_category='Secondhand Dealer - General' and license_status='Active'" --limit 3
agentdata company ch_six_fqs_search --segment EQ --pagesize 5
agentdata company eu_euronext_instrument_search --q "air liquide"
agentdata company it_borsa_quote --isin IT0003132476
agentdata company jp_jpx_stock_detail --qcode 7203
agentdata company jp_jpx_stock_list --section 1 --maxdisp 50

# economy (18)
agentdata economy bis_sdmx_series --flow WS_CBPOL --key M.US --start 2026-01-01
agentdata economy bls_series --series CUUR0000SA0,LNS14000000,CES0000000001 --start-year 2025 --end-year 2026
agentdata economy comext_flows --reporter DE --partner CN --product 85 --time 2024
agentdata economy ecb_fx_rates --currency USD --start 2026-10-01 --end 2026-10-31
agentdata economy eurostat_table --table nama_10_a64 --geo DE --time 2023
agentdata economy fred_series --series GDP --start 2024-01-01 --end 2026-11-01
agentdata economy ilo_stat_indicator --indicator EMP_TEMP_SEX_ECO_NB_A --ref-area USA --time-from 2020 --time-until 2024
agentdata economy imf_sdmx_series --flow WEO --country USA --indicator NGDP_RPCH --start 2023 --end 2026
agentdata economy oecd_sdmx_series --flow OECD.ECO.MAD,DSD_EO@DF_EO,1.5 --key DEU.. --start 2024 --end 2027
agentdata economy un_comtrade_trade --reporter USA --period 2023 --partner DEU --cmd TOTAL --flow X
agentdata economy unsd_sdg_indicators --search poverty
agentdata economy unsd_sdg_series --series DC_ODA_POVLG --geo China --time 2020
agentdata economy wb_countries --per-page 300
agentdata economy wb_ids_indicators --search "external debt"
agentdata economy wb_wdi_indicators --source 2 --search GDP
agentdata economy wb_wdi_values --country CHN --indicator NY.GDP.MKTP.CD --date 2020:2023
agentdata economy who_xmart_indicators --search tuberculosis --top 20
agentdata economy who_xmart_values --indicator TB_ppm_enrol_tx_prct --geo PAK --year 2023

# esg (8)
agentdata esg eiti_companies --name Equinor
agentdata esg eiti_payments --name Equinor
agentdata esg climatetrace_owners --name Gazprom
agentdata esg climatetrace_sources --owner-ids E100000001492 --year 2023 --limit 20
agentdata esg epa_frs_facilities --state-code DE --city-name MILTON --size 3
agentdata esg eutl_datasets
agentdata esg mx_sgm_concesiones --titulo 214536
agentdata esg uk_fhrs_establishments --name "pret a manger" --address london --pagesize 3

# healthcare (34)
agentdata healthcare rxnav_rxcui_by_name --name atorvastatin
agentdata healthcare rxnav_rxcui_atc --rxcui 83367
agentdata healthcare mesh_descriptor --label aspirin
agentdata healthcare gudid_devices --di 10198459305313
agentdata healthcare eudamed_devices_udi_di_data --primary-di 00191506004682
agentdata healthcare epmc_search --search 'AFF:"AstraZeneca"' --page-size 40
agentdata healthcare ctgov_studies --query.term 'AREA[LeadSponsorName]("Pfizer")' --page-size 3
agentdata healthcare ctgov_studies --query.titles aspirin --agg-filters phase:3 --page-size 3
agentdata healthcare dailymed_spls --search atorvastatin
agentdata healthcare fda_drug_shortages --search 'openfda.manufacturer_name.exact:"Fresenius Kabi USA, LLC" AND status:Current' --limit 3
agentdata healthcare fda_drug_recalls --search 'recalling_firm.exact:"SV Labs Toronto Corporation"'
agentdata healthcare fda_device_recalls --search 'recalling_firm:"Sedecal"' --limit 5
agentdata healthcare fda_food_recalls --search 'classification:"Class I" AND status:Ongoing' --limit 5
agentdata healthcare fda_crl_letters --search 'company_name:"InnoPharma"'
agentdata healthcare fda_drug_approvals --search 'sponsor_name:"TEVA"' --limit 5
agentdata healthcare fda_drug_catalog --search 'labeler_name:"pfizer"' --limit 5
agentdata healthcare fda_orangebook --search 'products.application_full_name:"Teva"' --limit 5
agentdata healthcare fda_device_510k --search 'applicant:"Medtronic"' --limit 5
agentdata healthcare fda_device_pma --search 'applicant:"Medtronic"' --limit 5
agentdata healthcare fda_device_registrations --search 'registration.name:"PLANET (ANHUI) INTERNATIONAL"' --limit 5
agentdata healthcare fda_device_classification --search 'device_name:"syringe"' --limit 5
agentdata healthcare gb_mhra_alerts --q Abbott --count 3
agentdata healthcare gb_mhra_alert --slug class-2-medicines-recall-xeal-pharma-clopidogrel-oral-liquids-all-strengths-el-26-a-slash-43
agentdata healthcare au_tga_artg --query '"Sun Pharma ANZ Pty Ltd"' --limit 5

agentdata healthcare au_tga_manufacturers --state QLD
agentdata healthcare br_cnes_facilities --codigo-uf 35 --limit 5
agentdata healthcare ca_dpd_companies --limit 5
agentdata healthcare ca_dpd_drug_products --status 3 --limit 5
agentdata healthcare ca_mdall_companies --limit 5
agentdata healthcare ca_mdall_licence --licence-id 111577
agentdata healthcare rasff_notifications --subject aflatoxin --items-per-page 5
agentdata healthcare safety_gate_alerts --year 2026 --limit 5
agentdata healthcare safety_gate_report --report-id 10000329 --limit 5
agentdata healthcare us_npi_organizations --organization-name "Mayo Clinic" --state MN --limit 3
agentdata healthcare jp_navii_facilities --q 内科

# legal (87)
agentdata legal afdb_debarred_entities --name "CHINA RAILWAY" --limit 3
agentdata legal ca_asc_enforcement --q Thomson --limit 5
agentdata legal ca_bcsc_disciplined --q Posen --limit 3
agentdata legal ca_cbsa_ruling --rulings-id 8000001670
agentdata legal ca_cbsa_rulings --limit 5
agentdata legal ca_sema_sanctions --country Syria --limit 3
agentdata legal ch_efk_party_financings --lang de --year 2025
agentdata legal ch_finma_enforcement_news --q "Julius Bär"
agentdata legal co_sic_sanctions --limit 3 --order "fecha_emitida DESC"
agentdata legal csl_screening_list --name "VTB BANK" --limit 3
agentdata legal de_lobbyregister_entries --identity ORGANIZATION --limit 5
agentdata legal ecfr_search --q asbestos --limit 5
agentdata legal ecfr_titles
agentdata legal dgcomp_cases_search --name Google --instrument SA
agentdata legal dk_ft_inspektionsredegoerelser_og_afgoerelser --year 2025 --page 1
agentdata legal idb_debarred_entities --q SHANDONG --limit 3
agentdata legal lda_filings --client-name "Google LLC" --limit 25
agentdata legal gb_ec_donations --query Sainsbury --limit 3
agentdata legal gb_ec_loans --query bank --limit 2
agentdata legal sipri_trades_search --supplier "United States" --type delivered --limit 3
agentdata legal eu_cijeu_search --q "state aid"
agentdata legal federalregister_documents --q asbestos --startdt 2026-01-01 --enddt 2026-09-30 --limit 3
agentdata legal courtlistener_opinions_search --q asbestos
agentdata legal fr_amf_sanction --theme all
agentdata legal fr_amf_transaction --theme all
agentdata legal gb_cma_cases --q google --count 5
agentdata legal gb_cma_case --slug nexfibre-slash-substantial-merger-inquiry
agentdata legal gb_tna_caselaw_search --party Barclays --per-page 3
agentdata legal gb_tna_caselaw_judgment --document-uri ewhc/admin/2026/2298
agentdata legal wb_debarred_firms --name "CHINA RAILWAY" --country CN --limit 3
agentdata legal gb_gazette_notices --insolvency --text "acme" --page-size 5
agentdata legal gb_nmw_round --slug 6-million-repaid-to-workers-as-government-cracks-down-on-employers-underpaying-their-staff
agentdata legal gb_nmw_rounds --count 5
agentdata legal ie_revenue_defaulters --part 1
agentdata legal us_occ_enforcements --year 2026 --month October
agentdata legal cfpb_complaints --search-term Discover --field company --size 3
agentdata legal cfpb_suggest_company --text Discover
agentdata legal in_indiacode_search --q "companies act"
agentdata legal in_sebi_orders --search Adani
agentdata legal tr_spk_fines --name "FERUDUN" --year 2025
agentdata legal tr_spk_bans --type company
agentdata legal ru_cbr_warning_list --violation pyramid --limit 5
agentdata legal ru_cbr_warning_list --site t.me --type internet_project --limit 5
agentdata legal br_bacen_regimes --cnpj 33171211
agentdata legal br_bacen_complaints --name ITAU --limit 5
agentdata legal br_bacen_complaints --tipo consortiums --name ADEMICON
agentdata legal br_cade_pesquisa_avancada --name PETROBRAS --limit 5
agentdata legal at_ris_judgments --court Justiz --case-number "2 Ob 52/16k" --limit 3
agentdata legal at_ris_statutes --q GmbH --limit 3
agentdata legal au_frl_titles --collection Act --top 5
agentdata legal ee_erjk_quarterly_reports --party-id 162 --limit 5
agentdata legal ee_erjk_election_reports --report-id 494641 --limit 10
agentdata legal es_boe_sumario --date 2026-11-10 --limit 5
agentdata legal ie_sipo_lobbying --query-text energy --limit 5
agentdata legal ua_nazk_parties --limit 5
agentdata legal ua_nazk_party_reports --party-id d0f1e224-e3de-45a2-8d12-43e4e9faee92 --limit 5
agentdata legal us_ferc_elibrary --docket CP17-101 --limit 5
agentdata legal fr_adlc_sanctions --id-decision 26-D-05 --limit 5
agentdata legal it_agcm_dettaglio --uid 0636576186F0C551C1258D5B0054C6E8 --replica dbconc
agentdata legal it_agcm_rating --words Telecom --maxresults 5
agentdata legal it_agcm_ricerca_provvedimenti --words Telecom --maxresults 5
agentdata legal it_arera_decisions --anno 2022 --numero 312
agentdata legal us_cbp_ruling --ruling-number 954699
agentdata legal us_cbp_stat
agentdata legal us_cftc_dispositions --page 0
agentdata legal us_cftc_enforcement --year 1995 --page 260
agentdata legal us_cftc_opinions --page 0
agentdata legal us_cftc_redlist --page 0
agentdata legal lv_knab_parties --limit 100
agentdata legal lv_knab_payments --party-public-id <public_id>
agentdata legal mx_cnv_sanctions --page-size 3 --order-field imposedOn --order-type DESC
agentdata legal mx_sabg_sanctioned_servants --q Abuso --limit 3
agentdata legal nl_rechtspraak_search --q onrechtmatige --page-size 3
agentdata legal pl_uokik_decision --unid EA4EB25AA6D1956FC1258EA3003605A7
agentdata legal pl_uokik_decisions --year 2026 --limit 5
agentdata legal kr_law_precedents --q 부가가치세 --page-size 3
agentdata legal kr_law_statutes --q 상속세 --page-size 3
agentdata legal fr_court_judgments --year 2026 --month 10 --court ce
agentdata legal fr_court_judgment --doc-id ECLI:FR:CECHR:2026:503183.20261001
agentdata legal us_occ_ea_search --q "Wells Fargo" --start-dte-min 2020-01-01
agentdata legal nl_acm_publications --category kartel --limit 5
agentdata legal nl_acm_publication --doc-id 896
agentdata legal es_cnmv_sanctions --party Sapiens
agentdata legal kr_fss_sanctions --layer institution --limit 5
agentdata legal au_accc_actions --limit 5
agentdata legal in_cci_antitrust_orders --q electricity --limit 5
agentdata legal in_cci_combinations --status Approved --form-type I --limit 5
agentdata legal jp_jftc_decision --case-id R080220R07G05000194K
agentdata legal jp_jftc_decisions --subject ソニー --limit 5

# procurement (7)
agentdata procurement br_pncp_contracts --data-inicial 20260901 --data-final 20260930 --pagina 1
agentdata procurement co_secop_contracts --where "upper(nombre_entidad) like '%MINISTERIO%'" --limit 5
agentdata procurement wb_procnotices --rows 5
agentdata procurement ny_abo_contracts --where "upper(vendor_name) like '%DELTA%'" --limit 5
agentdata procurement ny_mta_procurements --where "upper(vendor_name) like '%SKANSKA%'" --limit 5
agentdata procurement ungm_contract_awards --supplier DELTA
agentdata procurement ny_ogs_payments --where "upper(vendor) like '%HIGHLAND%'" --limit 5

# research (10)
agentdata research nih_reporter_projects --org-name "Stanford University" --fiscal-year 2024 --limit 3
agentdata research nih_reporter_publications --appl-ids 10491636
agentdata research nsf_awards --keyword "rare earth" --rpp 3
agentdata research openalex_institutions --search pfizer --per-page 3
agentdata research openalex_works --search aspirin --per-page 3
agentdata research crossref_works --funder 10.13039/100000001 --award-number OAC-1730655
agentdata research gb_gtr_projects --q lithium --s 10
agentdata research icite_pubs --pmid 26017442
agentdata research biorxiv_details --from-date 2026-10-01 --to-date 2026-10-07
agentdata research opencitations_citations --doi 10.1038/nature12373

# security (4)
agentdata security cisa_kev_catalog --limit 5
agentdata security euvd_exploited --limit 5
agentdata security euvd_vulnerability --id CVE-2024-3400
agentdata security osv_vulns --package lodash --ecosystem npm

# transportation (9)
agentdata transportation ca_ccar_aircraft --id 538042
agentdata transportation ca_ccar_search --owner "AIR CANADA" --limit 3
agentdata transportation euasl_datasets
agentdata transportation paris_mou_bans
agentdata transportation paris_mou_detentions
agentdata transportation paris_mou_inspections --period-min 01/12/2025 --period-max 31/12/2025 --ism-company-imo-number 0052193
agentdata transportation paris_mou_low_performance_companies
agentdata transportation thetis_mrv_datasets
agentdata transportation us_fmcsa_carriers --where "dot_number='963305'" --limit 3

Caching

Every response is cached in a per-user SQLite cache shared across processes — cache hits cost no rate budget, and entry lifetimes follow the per-source TTL table in the rate-limits ledger; immutable lookups (e.g. drug name → RxCUI) are cached permanently. A fresh install in a new virtual environment will reuse entries written by earlier runs on the same machine — that is by design.

Default %LOCALAPPDATA%\agentdata\cache.db (Windows) · ~/.cache/agentdata/cache.db (macOS/Linux)
Override AGENTDATA_CACHE_DIR env var · --cache-dir (CLI)
Bypass / clear --no-cache (CLI) · no_cache=True (Python) · delete the file

Documentation

Releases

The complete batch log — every release back to 0.1.0 — is kept in the Chinese README.md. The most recent batches:

0.3.68 — Batch E10, the developed-economy exchange faces: five listed-universe/quote faces (+5 faces, 251→256; 169→173 providers; 35 jurisdictions held; company 74→79 — first Italian company face, zero new jurisdictions). ch_six_fqs_search — the SIX Swiss Exchange FQS command API (column oracle pinned: invalid select names are 400-rejected verbatim, so the engine refuses them locally; price columns are opt-in under SIX's commercial-use caveat). eu_euronext_instrument_search — the live Euronext venue resolver: one ISIN lists on many MICs, the join key is ISIN+MIC, and the zero-hit sentinel row is dropped rather than counted. it_borsa_quote — the Borsa Italiana dati-completi card by ISIN (market cap parsed from Italian number format; an unknown ISIN or mismatched MIC serves a generic page silently — reported found=false, never a fabricated card). jp_jpx_stock_list + jp_jpx_stock_detail — the JPX qjsonp gateway behind its Referer door: the 4,490-name listed universe page by page (board + 33-sector industry on every row) and the 57-field issue card (ISIN, market cap JPY, shares, embedded daily OHLCV, EPS history). Two same-day upstream flips deferred with probe evidence: Wiener Börse (TLS serving a *.factsetdigitalsolutions.com certificate on both hosts) and TMX (tsx.com flipped to an AWS WAF JS challenge between the morning's green probes and the afternoon smoke — a urllib client cannot execute the challenge; re-probe later).

0.3.67 — Batch E9, the reliability fix: us_sec_fulltext_search --census (+0 faces, 251/169/35 held — a mode on an existing face). The plain mode returns one 100-hit page and post-filtering it systematically undercounts; census sweeps the whole window — date bisection until every leaf window is exact, every page harvested, deduped by filing accession. --census-entity is the local whole-word entity filter (the upstream does not parse company:"..." — the quoted prefix is just a search term; "IREN" never hits "Siren"); a single-day window still over the 10,000 floor is reported in overflow_windows, never dropped; the request budget (default 400) fails loudly with progress. Reference-date anchors, all complete and equal to manual windowed re-runs: IREN×Nvidia 87 filings (50 current-name + 37 old-name Iris Energy — one issuer), CoreWeave×OpenAI 37 (34 issuer + 3 co-filed 425), APLD×Nvidia 54 (the hand count of 43 had dropped the S-1 registration family) — strict supersets with transparent composition.

0.3.66 — Batch E8, the deferral reopening: the SEC XBRL structured-facts family (+4 faces on one new provider, 247→251; 168→169 providers and 35 jurisdictions held; company 70→74 — US deepened only).

One keyless provider edgar_xbrl on data.sec.gov, live-probed 2026-11-17 — the quantitative layer under the edgar_cik registry layer, reopening the 2026-10-19 deferral (the local XBRL bulk that ruling leaned on was lost with the E: drive, and the MCP/OmniHub surface carries no financial-data paths). us_sec_companyconcept is the cheap server-side point query (18 KB vs the 4 MB dump), us_sec_companyfacts is the issuer dump in inventory mode (taxonomy → concept-name list — the drift-discovery face) XOR concept mode, us_sec_frames answers one concept × unit × period cross-section for the whole market (NetIncomeLoss CY2024 = 6,084 entities), and us_sec_tickers is the ticker↔CIK directory riding an ABSOLUTE path to www.sec.gov (one provider id, two hosts, each with its own polite bucket; exactly one selector — bare enumeration refused, rows 10,404 ↔ 7,997 CIKs because share classes share a CIK). Three live laws baked in: the CONCEPT-DRIFT LAW (us-gaap tags drift across taxonomy versions — Revenues is dead after 2018 → RevenueFromContractWithCustomerExcludingAssessedTax; issuers simply lack concepts — Apple has no EarningsPerShareDiluted; tags pass through verbatim, absence is a legal answer); the FY/FP LAW (a fact's fy/fp stamp the FILING, not the fact period — Apple's FY2026 10-K carries the FY2025 comparison period stamped fy=2026; period membership reads off start/end); and the FRAME LAW (CY / CYQ / CYQI only). Rate 5/5 — edgar_cik 5 + edgar_xbrl 5 = 10 ≤ the official 10 req/s with a declared UA (the §7 exception register's third member); TTL companyfacts/companyconcept/frames 7d, tickers 24h; rate-limits v1.54 (169 providers / 251 faces); pinned by tests/test_flat_0366.py.

0.3.65 — Batch E7, single-face deepening: US CFTC OGC Opinions & Adjudicatory Orders (+1 face born flat, 246→247; 168 providers and 35 jurisdictions held; legal 86→87 — US deepened only).

One keyless face on the extended-in-place us_cftc provider, live-probed 2026-11-17. us_cftc_opinions reads the Office of the General Counsel's reparations-appeals register — the layer above the dispositions court: parties-first captions ("Eric Thompson, … v. …"), ≈87 decisions over 9 pages (0..8), date-descending 2012→2025. Three live laws baked in: the register anchors DUAL-ERA PDF paths (recent /sites/default/files/YYYY/MM/ plus the legacy /idc/groups/public/%40lropinionsandadjudicaryorders/documents/legalpleading/ kept verbatim — %40 is the encoded @ and "adjudicary" is the upstream's own misspelling, never normalized); multi-docket rows caption the plural "CFTC Docket Nos. 20-R007; …; and 20-R025" — the whole semicolon list captured as one docket_no string; and legacy-era docket captions scatter across tolerance shapes (the "CFTC " prefix dropped, non-R numbering "11-E-01", NFA arbitration pairs "CRAA 12-03"), while the two 2012 rows with no caption at all report None honestly. Same host browser-suite gate as the other three views; TTL 24h; rate-limits v1.53 (168 providers / 247 faces); pinned by tests/test_flat_0365.py.

0.3.64 — Batch E6, the negative-list batch: UK NMW naming-scheme pair + US CFTC RED List & reparation decisions (+4 faces born flat, 242→246; 167→168 providers; legal 82→86 — GB/US deepened only, zero new jurisdictions).

Two keyless providers, live-probed 2026-11-16. gb_nmw_rounds/gb_nmw_round read the government's "named and shamed" underpayment register: the discovery face pins the gov.uk search API to the scheme's own query (41 releases, Round 22 newest), and the dossier face fetches the press-release HTML because the content API answers an EMPTY body for these pages (attachments live only in the HTML) — the named-employers xlsx coordinates come out per round, with the no-year-suffix law (hand-assembling …_16_October_2025.xlsx 404s; the real file has no year). On the US side us_cftc_redlist/us_cftc_dispositions extend the CFTC provider in place: the RED List's 284 unregistered foreign entities are ALPHABETICAL with dates scattered (widening items_per_page answers zero rows — the views setting is decorative), and the reparation decisions anchor IS the decision PDF with the docket riding the blurb cell. Re-probes the same day: uk_ico's Cloudflare managed challenge now covers the main-site sitemap (was keyless 2026-10-30) and us_fcc still answers the 387B Access-Denied stub — both stay deferred. Polite tiers: gb_nmw 1 rps / us_cftc stays 0.5 rps (the two new ops ride the provider's existing tier); rate-limits ledger v1.52 (168 providers / 246 faces); compliance booklet v2.6 (🔴16/🟡23/🟢92/⚪37); pinned by tests/test_flat_0364.py.

0.3.63 — Batch E5 depth wave: JP JFTC decision pair + IN CCI register pair + AU ACCC actions feed (+5 faces born flat, 237→242; 164→167 providers; legal 77→82 — AU/IN/JP deepened only, zero new jurisdictions; Japan's first legal faces).

Three keyless providers, live-probed 2026-11-16. jp_jftc_decisions/jp_jftc_decision read the JFTC decision database behind a UA gate and the CSRF token/cookie pair dance (token scraped per session, both must ride together); the era tables (R/H/S → 2019-… mapping) and full-width event numbers are kept verbatim, and case_id is the hard key across the pair. The CCI registers pair (in_cci_antitrust_orders/in_cci_combinations) rides the DataTables POST channel — the GET twin answers 405 (the pair trap) — and tokens MISALIGN across the two registers, so column offsets never transfer; the XYZ placeholder party names ship verbatim (upstream's own masking), and combination rows keep their raw HTML soup byte-exact. au_accc_actions rides the lazy-views two-hop (GET /views/ajax with view_path=/node/162342, f[] array form, bare JSON command envelope, 0-based page): the rows window is WAF-gated — ~50 rapid calls close it (even an exact replay answers the 64KB shell) and it self-heals after ~25 minutes; the RSS feed was rejected (no topic/type filters).

0.3.62 — Batch E4 depth wave: NL ACM publications pair + ES CNMV sanctions + KR FSS sanctions (+4 faces born flat, 233→237; 161→164 providers; legal 73→77 — NL/ES/KR deepened only, zero new jurisdictions).

Three keyless providers, live-probed 2026-11-16. nl_acm_publications/nl_acm_publication read the Dutch competition authority's decision archive through the channel that actually scales: the list page's JS window does not widen (measured), so the SITEMAP is the channel — one list call sweeps the 9-page sitemap index plus all submaps, 13,401 URLs back to 2003; the category taxonomy is a slug-substring heuristic (word-boundary matching false-negatives compounds: kartelafspraken counts as kartel), and the detail page's Parties block rides a nested second <h3> inside the case content block. es_cnmv_sanctions serves the Spanish securities regulator's sanction register behind a Referer contract, and its five-year window SLIDES (19 pages pinned 2026-10-19 → 18 on 2026-11-16) — the tail page is always discovered from page-0 links, never hardcoded, and boe_ref rows join the ES BOE gazette face; it rides a second provider id beside the 0.3.44 company face (the wb/wb_proc precedent). kr_fss_sanctions is the tree's first POST × bytes combination (urlencoded excelDown POST → 270 KB xlsx, 5,804 rows) with the header-misalignment law: cells are LETTER-keyed in the sheet XML, never order-keyed — and 1,245 rows carry their substance in a PDF attachment, answering honest nulls, not fabricated values.

0.3.61 — Batch E3: FR administrative-courts judgment archives + US OCC enforcement database (+3 faces born flat, 230→233; 159→161 providers; legal 70→73 — FR/US deepened only, zero new jurisdictions).

Two keyless providers, live-probed 2026-11-16. fr_court_judgments/fr_court_judgment read the administrative-courts portal's MONTHLY zip archives at fixed directory paths (/DCE/{yyyy}/{mm} for Conseil d'État, /DCA/ for appeal courts) — a dead-looking dataset page is not a dead source: the archive zips stay live while the data.gouv dataset page lags years behind. The zip rides the bytes channel (an Accept wall that only takes */*) and unrolls one namespace-free XML per judgment via stdlib zipfile; the 2026-10 measured census was CE 791 / CAA 2,340. The CAA archives carry NO ECLI element (measured 0/24) — ecli answers None on CAA rows and the single-judgment face keys on the archive file stem, where a CE ECLI maps onto the same stem. A month with no zip answers upstream_unavailable, never a fabricated empty. us_occ_ea_search is the OCC's EASearch structured database — the 0.3.28 enforcement news face's twin under a second provider id (that face name was taken; the new face takes the upstream portal name). The FULL history 1989→2026 (~6,072 actions) answers UNFILTERED in one 1.2 MB shot, so the face refuses filter-less calls locally (bulk stays bulk); rows carry CharterNumber (the bank-register join key) and AA-ENF-YYYY-NN dockets — the same docket key the 0.3.28 face parses out of prose.

0.3.60 — Batch E2 rulings/registers cluster: AU FRL legislation titles + BR CADE cases + CA CBSA advance-rulings pair + DE Bundestag lobby register + DK FSA supervision archive + IT AGCM trio + PL UOKiK decisions pair + US CBP CROSS rulings pair + BR BNDES financing (+14 faces born flat, 216→230; legal 57→70, company 69→70; nine keyless providers).

The sixth deepen-only batch: the legal domain gains its first AU, DE, DK and PL faces — AU and DE pre-exist the ledger through other domains, but DK and PL are new to the ledger (33→35; the batch's own registry line claimed "zero new countries (33 held)" and was measured wrong — corrected here by the 0.3.62 audit). Highlights: the AU OData channel forbids percent-encoding the $-prefixed option names (%24top → 400); BR CADE answers every scalar base64-wrapped in mixed UTF-8/cp1252 and a no-match is an EMPTY answer, not a failure; CA CBSA rulings require lang=en (absent = 400) and answer the whole 1,234-ruling register in one shot; the DE lobby register takes the 18 MB keyless JSON export (the official REST is ApiKey-gated → off route under the keyless-only ruling); DK FSA requires a Referer header and a nested-object userInput; IT AGCM sits behind an IP-bound Cloudflare gate and refuses multi-word search tokens locally; US CBP CROSS details key on string ruling ids.

0.3.59 — Batch E1 enforcement/party-finance first slice (+8 legal faces born flat, 207→216; legal 49→57 — FR AMF sanctions/transactions pair, EU CJEU case-law search, LV KNAB party-finance pair, CH EFK party financings + FINMA enforcement news, IN India Code search; Latvia new to the ledger, 32→33).

Eight keyless faces across six providers, live-probed 2026-11-13. eu_cijeu_search reads the EU Court of Justice's Curia search; the FR AMF pair rides the sanction Commission's portal; LV KNAB carries Latvia into the ledger with the party-annual-report pair; CH EFK reads party financings and FINMA's enforcement-news feed; IN India Code opens the statute text line. All polite-tier keyless with no published quotas — walls are channel walls, not rate lines.

0.3.58 — Batch D same-stack widening: FDIC BankFind sibling quartet + FR INPI accounting-ratio pair + CA ASC enforcement + BCSC disciplined + BR CNES facilities (+9 faces born flat, 198→207; company 62→68, legal 47→49, healthcare 33→34 — healthcare's first BR face; ca_amf deferred at the verify gate, CF-403 on non-browser clients).

Same-stack widening of existing deepeners: the FDIC quartet rides the BankFind suite's sibling endpoints (failures / financials / history / locations), the INPI pair brings FR company accounting ratios (one of the hardest-to-reach French datasets, now keyless), and BR CNES opens healthcare's BR axis. The urlencoded channel gains doseq list→repeated-key encoding (inert to every existing scalar op).

0.3.57 — Batch C major-power axis II: ES BOE daily gazette table of contents + ES OEPM IP gazette index + TR KAP member/disclosure pair (194→198 faces, 136→139 providers; company 59→62, legal 46→47 — jurisdiction count unchanged at 31, existing countries deepened only).

Three keyless providers, live-probed 2026-11-12 (tmp/probe-sdk-0357/). es_boe_sumario reads one publication day's full BOE table of contents (www.boe.es/datosabiertos — the JSON channel rides content negotiation, which the SDK's default Accept header already sends; the same www.boe.es host bucket is shared with es_borme across domains): every container level arrives dict-OR-list (singleton-collapse) and items nest in four measured shapes, all collected with a seccion/departamento/epigrafe breadcrumb plus a depth-blind census that reconciles against the parsed rows — a drifted shape surfaces as a warning, not vanishing rows. No-publication days answer 404 XML and surface as published=false; the archive reaches back to the 1960s. es_oepm_bopi is the industrial-property gazette issue index (sede.oepm.gob.es/apibopiweb): the no-issue negative is a 200 with a ZERO-BYTE body (weekend AND pre-1995 both measured) — carried as "" through the accept:'text' channel and structured as published=false; the UCM artifact keys are the real keys (id/fechaPublicacion arrive null upstream by design). The tr_kap_members/tr_kap_disclosures pair covers the Turkish public disclosure platform (www.kap.org.tr — the gate measured open all four ways, bare client shipped): the member register carries the TR company triple taxNo+tradeRegNo+tradeRegOffice with the W-state fake placeholder row filtered and counted, and the disclosure stream caps at EXACTLY 2,000 rows per window with no total echo — likely_truncated flags it and the window is walled at 31 days locally. tr_ekap deferred at the verify gate (the human-verification gate came ON).

0.3.56 — Batch B major-power axis I: DE register search + JP IP trio + KR patents + JP NAVII facilities + KR statutes/precedents (186→194 faces, 131→136 providers; company 54→59, legal 44→46, healthcare 32→33 — jurisdiction count unchanged at 31, this batch deepens existing countries only).

Five keyless providers, live-probed 2026-11-12 (tmp/probe-sdk-0356/). de_register_search queries the German Unternehmensregister through a three-hop token chain (POST landing mints a 231-char token cached 45min → GET /de/suche with RSC:1 + Accept: text/x-component → a React server-component stream parsed by a string-aware brace matcher; the companyDto row's euid (DED2601V.HRB252581 — state + court + register type + number) is the strong join key, a bare HRB number needs its court code; searchTerm "exact" is a phrase match; the search hop is never cached). jp_trademark_search / jp_patent_search / jp_design_search break J-PlatPat with the client-side hash gate — the SPA serves no cookies, the client mints U/L/D/S/T itself and signs Z = sha256 over the ECRL bit scheme (ECRL=105 = 0b1101001 → bits 1,3,6 ride: sha256(U + "ja" + ECKY), bit4 D does NOT); one cached ECRL/ECKY pair boots all three faces; the year slice (FILTER_INFO ITEM/COND/COMPAR_COND_SIGN) is the enumeration key and the SIGN is mandatory (a bare COND is silently ignored — measured); 3,000-row per-call cap with no paging keys; AUTO_NARROW_FLG=1 gives free counting via the pre-narrow total; the patent face sends no OTID (a null is stripped by the transport and ≡ key absent, A/B-verified back-to-back 2026-11-12); RSLT_CD=8 is the captcha gate — raise, never retry through it. kr_kipris_patents reads KIPRIS (form-urlencoded GET; AN+APC row keys, TL keeps the <!HS>/<!HE> highlight anchors verbatim, "98" = zero hits, page cap 255 / page-size cap 2000). jp_navii_facilities walks the MHLW NAVII statutory facility directory through a three-hop session chain (initialize mints the AWSALB/SESSION jar cached 20min → iryoSearch with the ajax headers returns the session-bound 検索ID, never cached → S2400 with 0-BASED pages of true 20-row pagination; keyword_type 3=診療科目 enumerates 内科 59,281 live; pharmacy=True rides initializeYakk/yakkyokuSearch onto the same S2400 page, kikanKbn=5; ~105k 医療機関 + ~61k 薬局). kr_law_statutes / kr_law_precedents read law.go.kr (keyless urlencoded POST; the HTML fragment IS the payload — accept:"text", the nl_rechtspraak precedent; statutes ride the lawNm form family with the efYd 99990101 pending sentinel, live fragments echo NO total and none is fabricated; precedents carry the [대법원 date 선고 caseNo 판결] metadata on javascript: anchors). Rate tiers: de_register/jp_navii 0.3 rps, jp_jplatpat/kr_kipris/kr_law 0.25 rps (the 4s cadence is the captcha discipline); rate-limits v1.44 (136 providers / 194 faces); offline pins tests/test_flat_0356.py (39 cases — it caught four real bugs: the missing accept:"text" on kr_law ops, the missing cache-payload unwrap in jp_navii._session, the missing no_cache parameter on the jp IP face helpers, and the DE reason/evidence reading the wrong registerNumber key). Verified: full offline suite + one live shot per face (2026-11-12).

0.3.55 — MX legal pair + NL Rechtspraak + MX SGM mining cadastre (182→186 faces, 131 providers; legal 41→44, esg 7→8 — the SDK's first NL legal and first MX legal/esg jurisdictions, 30→32). Four keyless providers, live-probed 2026-11-12. mx_cnv_sanctions reads the CNBV financial-sanctions register through Azure APIM — keyless by publication (the subscription key ships hardcoded in the public React bundle, the AU ACMA family; the OpenAI key leaked in the same bundle is someone else's credential and is deliberately not carried). The pagination lives in the RESPONSE headers — the body is a bare row list and x-total-items/x-total-pages carry the totals, which is why transport extension 11 (resp_headers) exists; the entire paging state rides request headers (X-Page-Number 0-based, X-Page-Size, X-Order-Field/X-Order-Type — imposedOn verified both directions), so the op is never cached by design: the URL is identical for every page and a cache replay carries no headers (totals answer None, never fabricated). 11,593 rows live. Coverage since the 2014 Lineamientos INCLUDES the non-regulated AML faces (centros cambiarios / transmisores de dinero / SOFOMES E.N.R.) that no other MX register carries; obligatedSubject is the company join key. mx_sabg_sanctioned_servants rides the CKAN datastore bypass (the direct CSV is Akamai-403 on every stack; the datastore inlines the whole 435-row in-force register as JSON): the 13 statutory columns verbatim with inicio→fin the live barring window; q is full-text (Abuso → 78), filters JSON exact-match (LFRASP → 364). nl_rechtspraak_search is a strict-shape POST (~479k published cases, the register reaches 1847): every PascalCase body key must ride and Field must stay zt0 or the upstream answers 200 with an Angular error page instead of JSON; TitelEmphasis IS the ECLI, CorrelationId is the deterministic md5 digest of the request itself so the cache key stays stable, and pdf_url is derived per row without fetching. mx_sgm_concesiones reads the SGM mining-concession cadastre (ArcGIS FeatureServer layer 49, 19,840 concessions): ctithomolo the holder-company join key, an at-least-one-filter wall (a bare scan is batch-pipeline territory), and inicio_de_ an Excel-base epoch-ms whose NEGATIVE values are legal (titulo 76659 → 1829-12-28).

0.3.54 — transportation domain chartered (ninth domain, 180→182 faces, 127 providers) — seven faces migrated prefix-only + Canada TC CCAR pair new. Per the 0.3.54 ruling the transport-compliance faces move out of esg/company in one cut: euasl_datasets, the four paris_mou_* faces and thetis_mrv_datasets leave esg (13→7 — esg is environment + social again), us_fmcsa_carriers leaves company (55→54 — a USDOT number is a carrier-census key). Names, parameters and jurisdictions are unchanged; only the command prefix moves to agentdata transportation …, and a wrong-domain call raises usage_error naming the right domain (the guard fires at call time). Two new faces: ca_ccar_search reads Transport Canada's Civil Aircraft Register (CCARCS-RIACC — the POST form's 302 lands on a stateless bare GET RchSimpRes.aspx, a five-slot contains pipeline (cn/mn/sn/on/m) wrapped with print=y answering one full page: AIR CANADA 269 rows / 365KB live; contains is case-sensitive against the register's stored form, pinned live 2026-11-11 — owner="Air Canada" answers 0 rows, owner="AIR CANADA" answers 269; total_matched is the honest upstream total and limit 1-500 is a local cap; at least one slot is required locally; adet_id is the join key into the aircraft face). ca_ccar_aircraft reads the single-aircraft dossier (ADet.aspx?id=) — four blocks (aircraft / manufacturer / registration / registered owner) flattened into one dict; unmapped labels land snake_case in extra_fields rather than being dropped (mail_recipient proven); the ~500k id space has holes — a not-found is a legal empty, and ids come from search rows, never guessed; natural-person owner names and addresses are recorded verbatim (us_fmcsa company_officer_1/2 public-register precedent; the raw harvest line strips contact details — domain red line). Both keyless, no gate and no WAF (probe tmp/probe-ca-ccar/ 2026-11-10); ca_ccar at 0.5/2; rate-limits v1.42 (127 providers / 182 faces) + compliance register v1.6 (canada.ca GC umbrella terms 🟡: non-commercial reuse is free, commercial redistribution needs prior written permission); command-naming rows 181-182; offline pins test_flat_transportation.py. Verified: full offline suite plus one live call per new face (2026-11-11).

0.3.53 — FR Regafi pair + UK FHRS + US NPPES (+4 faces, 176→180; company 53→55, esg 12→13, healthcare 31→32). Four keyless providers, probes pinned 2026-11-10 under tmp/probe-sdk0352/. fr_regafi_entities / fr_regafi_authorisations read the ACPR's official financial-sector register (regafi.fr — the React shell's real backend is the Opendatasoft Explore API v2.1): a 25,890-row supervised-entity roster (SIREN + LEI dual hard keys, the parent edge rides every row, nom_commerciaux carries renaming history as JSON) and its 31,107-row per-authorisation timeline (date_debut runs 1946→today; the ~40 *_json columns carry per-service authorisation blocks with their own effective dates — ODS serialises them sometimes as strings, sometimes pre-parsed, and the face normalises in place). uk_fhrs_establishments searches the FSA Food Hygiene Rating Scheme live (~615,098 establishments; the x-api-version: 2 header is the only gate; postcode is 403-walled as a query param on this endpoint — the PostCode field rides every row instead). us_npi_organizations reads the NPPES registry pinned to NPI-2 organizational records: org-name search is exact-phrase ("Mayo" answers zero, "Mayo Clinic" answers 2), limit>200 is silently clamped upstream (measured: 250 → result_count 200) so 1-200 is a local wall, and NPI-1 rows still mix in under enumerationType=Organizational — filtered client-side with the result_count/count pair shipped honestly.

0.3.52 — discovery surface machinery batch (0 new faces, 176 unchanged). agentdata capabilities renders the complete face table — every realtime face with its one-line summary, jurisdiction, execution plane, key requirement, and per-parameter closed sets — straight from the argparse command tree, so it cannot drift and ships with every release for free (--format json|md, --domain-filter, --face; runs offline, no PAT). Three CLI repairs ride along: 32 faces in economy/research/security regained their help= catalog lines (their domain --help used to list nothing), the verb layer takes metavar=FACE (the legal usage line no longer spells 38 face names), and allow_abbrev=False now covers the whole tree (a mistyped --api-k used to silently become --api-key). Cache keys are pinned to the provider's base URL (cache_key_for(scope=...)): a provider id repointed at another upstream — Comext's dedicated root vs plain Eurostat — can no longer answer from the old host's cache entry. Offline pins: test_capabilities.py (20 tests); full suite 1497 passed. The server-side capability document is a P2 byproduct — three clauses recorded in docs/domains/api.md §9.7.

0.3.51 — legal enforcement tier-two trio + company FMCSA carrier census (+4 faces, 172→176; legal 38→41, company 52→53). The SDK's first IT jurisdiction (29→30; FR enters the legal domain for the first time — fr_recherche lives in company); four keyless providers (fr_adlc / it_arera / us_cftc / us_fmcsa), probes pinned 2026-11-09 under tmp/probe-sdk0350/l5/. fr_adlc_sanctions reads the French competition authority's company-fine ledger through its official data.gouv.fr mirror (the authority's own domain egress-blocks this exit) via a two-hop chain: the datasets API resolves the latest CSV resource, whose build-stamped URL rolls on every update (dataset TTL 1h + CSV TTL 24h; transport extension 10 absolute-URL fetch). 1,208 rows × 5 columns; amounts arrive as decimal strings and cast to float locally; case total ≠ sum of individuel rows — both figures are upstream-published and shipped verbatim, never reconciled (26-D-05: rows sum 15,260,000 vs the register's own 12,670,000; 15-D-19 ships €672,332,000 case-level). it_arera_decisions reads the Italian ARERA regulated-sector decision library on bare TYPO3 — no gate, no WAF, 15,599 decisions back to 1996 with the count living in the pager; tipologia is a nine-word exact-match closed set, and the PDF path is never assembled from the sigla (fileadmin naming drifts by era — the face ships the detail anchor). us_cftc_enforcement reads the CFTC enforcement-actions Drupal views table — the one live surface (solr sits behind Cloudflare, publicreporting is a resold domain, data.cftc.gov resets) — with a UA gate that drifted the same day (bare UA passed in the morning probe, 403'd in the afternoon one; the ext-8 browser suite is permanent). Pages are the epochs: ?page=0..260 ends at the register's 1995 floor (~2,610 actions), 1998 is absent from the register's own year filter and the wall records it as-is. us_fmcsa_carriers reads the FMCSA Company Census File (data.transportation.gov az4n-8mr2, the tx_tdlr SoQL-verbatim contract): 4,528,322 rows of FULL history ($select=count(*) — the ~400k audit figure was the active face, an order of magnitude off); Socrata omits null fields per row (53 keys on a populated row vs ~58 catalog columns); dot_number is the CSA/SMS join key and dun_bradstreet_no is the rare official DUNS cross-key ("0" sentinels kept verbatim, suppressed from evidence quotes); sibling set 4a2k-zf79 is a 403 dead single-dataset — az4n is the live set. fr_adlc/it_arera/us_cftc at 0.5/2, us_fmcsa at 1/2; rate-limits v1.40 (123 providers / 176 faces); offline pins test_flat_legal_0351.py + test_flat_company_us_fmcsa.py. 0.3.50 — legal enforcement/party-assets realtime sextet (+6 faces, 166→172; legal 32→38). The SDK's first EE and UA jurisdictions (27→29); four keyless providers (us_ferc / ie_sipo / ee_erjk / ua_nazk), probes pinned 2026-11-09 under tmp/probe-sdk0350/. us_ferc_elibrary is a POST AdvancedSearch behind a Cloudflare UA-fingerprint gate (polite agentdata UA = 403; the browser-suite header set passes) — totalHits 4,722,924 back to 1980, resultsPerPage ceiling 1000 verified, and the upstream's own triple-s acesssionNumber spelling is recorded verbatim as the official key (TTL 12h). ie_sipo_lobbying takes 13 mandatory params where an explicit empty string is required (one missing = 404), dates are DD-MM-YYYY (ISO = 500), Total = 123,591, and Lobbyist.CompanyRegistrationNumber is the CRO join key. The ERJK pair rides the erjk.ee /et/ prefix wall (bare /api is a 301): the quarterly-receipts matrix (455 rows) plus the party register (18) and person-keyed report bodies (reg_code = the RIK key). The NAZK pair is all-POST on politdata (GET = 404) with a server-side pageSize clamp (10→20 — the local default is 20 and never pretends otherwise) and public_summary embedded per report. All four at the 0.5/2 polite tier; rate-limits v1.39 (119 providers / 172 faces); offline pin test_flat_legal_0350.py. 0.3.49 — company US city-license Socrata trio (+3 faces, 163→166; company 49→52). chi_bacp_licenses reads the Chicago business-licensing register (data.cityofchicago.org r5kz-chrr: 1,210,956 rows, full history back to 1997), lacity_registrations reads the LA City register (data.lacity.org 6rrh-rzua: 632,799 rows), and nyc_dcwp_licenses reads the NYC DCWP consumer-affairs license book (data.cityofnewyork.us w7w3-xahh: 72,452 rows). All three are SoQL verbatim channels — the co_rues contract — and the state-axis ruling of 0.3.48 extends to the city axis: faces are named on the host/locality stem, and lacity is deliberately not la_ (LA is ISO-3166 Laos; SDK prefixes are country codes). Chicago refreshes daily via payment_date (TTL 24h); NYC's book lags ~2.5 months (TTL 7d); the LA table is a registration book, not a license book. Socrata system columns (:@computed_region_*) pass through verbatim — and drift per row on the LA host (5↔6), while the NYC host ships none. 0.3.48 — company US state-license Socrata twin (+2 faces, 161→163; company 47→49). tx_tdlr_licenses reads the Texas TDLR all-licenses register (data.texas.gov 7358-krk7: 966,778 rows × 26 view columns — 18 real + 8 Socrata :@computed_region_* system columns) and wa_doh_credentials reads the Washington DOH health-care credential register (data.wa.gov qxh8-f4bd: 2,468,860 rows, refreshed daily — the count moved +1 past the research probe the same day). Both are SoQL verbatim channels (raw $where, $limit 1-500, $order without NULLS, no total echo and none faked) — the co_rues contract — and the SDK's first US sub-national state-axis faces, named on state-code prefixes. Socrata omits null fields per row, so a bare individual row ships 9 keys while an establishment row carries the whole business-address family — the individual-vs-establishment duality lives in the data, not the schema. The WA birthyear column is stripped from every row (the PII red line: birth year + full name is exactly the combination the house rules rule out — names stay, birth years do not). 0.3.47 — legal AT RIS OGD pair (+2 faces, 159→161; legal 30→32). at_ris_judgments / at_ris_statutes read the Austrian federal legal information system's open-data API (data.bka.gv.at, OGD API v2.6) — the SDK's first AT jurisdiction: sixteen Judikatur shards (1,008,601 documents; Vwgh alone 357,857 and drifting daily) plus the consolidated Bundesrecht (442,058 documents). The gate is a parameter wall, not a rate line: Applikation is case-sensitive (VWGH is answered by an HTTP 200 whose error body self-exposes the schema dictionary, and the boolean-flag form is silently ignored and quietly answers the Justiz shard — the deadliest of the three postures, so an exact-case 16-value closed set sits before the wire), page size is an upstream constant (20), and beyond the last page the upstream answers HTTP 500 rather than an empty page.

  • The code in this repository is licensed under the license in LICENSE.
  • The data returned by the SDK is governed by the terms of each upstream source; the SDK grants no data rights. See TERMS.md.

Contact

Metadata

Release files for agentdata 0.3.68

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for agentdata 0.3.68
File Size Uploaded Provenance Actions
agentdata-0.3.68.tar.gz 856.1 kB With provenance Details for agentdata-0.3.68.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for agentdata 0.3.68
File Interpreter ABI Platform Actions
agentdata-0.3.68-py3-none-any.whl Python 3 none any Details for agentdata-0.3.68-py3-none-any.whl

Total release size: 1.8 MB

Release files / agentdata-0.3.68.tar.gz

Download URL agentdata-0.3.68.tar.gz
Size 856.1 kB
Tags Source
SHA-256 checksum
How to use checksums
ec8f9860e081ba027835affcc7c615704e4baf9de4a2e5da29d9325b2b7e95d0
BLAKE2b-256 checksum
How to use checksums
ab18e0d30423d52650ee62d9432f73f48dedb09550d8904804996e42fdad6142
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 18, 2026.

Transparency log

Release files / agentdata-0.3.68-py3-none-any.whl

Download URL agentdata-0.3.68-py3-none-any.whl
Size 950.1 kB
Tags Python 3
SHA-256 checksum
How to use checksums
b77c41001af606593de283dc15206dfcd14ab84d43a0ceec00991f14170812bc
BLAKE2b-256 checksum
How to use checksums
9288fdf41ad0ad1fcfc066de35bf7a1517e6b48ae9479c68c0fbeaea03d0e897
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 18, 2026.

Transparency log

Security for agentdata 0.3.68

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for agentdata-0.3.68.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 18, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page