KYB Company Verifier - Cross-Registry Business Verification API

Verify a company name against eight official, keyless company registries in one call and get, per registry: whether a record exists, the normalized legal status, jurisdiction, identifiers, formation date, registered agent, officers, and a graded name match.

This is a screening and grounding tool over public government registries. It is not a KYB, AML or compliance determination. A registry record is evidence, not a decision. Confirm identifiers against the source registry before onboarding, contracting or paying.

The rule this actor now enforces: a partial check is never presented as a complete one

Three things could previously make a partial verification look like a finished one. All were reproduced live and fixed on 2026-08-01.

1. A registry that failed was skipped entirely - no row, no field, nothing. You got seven rows instead of eight and no way to tell which registry was never checked. If every registry failed, the run emitted zero rows and reported SUCCESS: a silent, confident "we found nothing". Now every unreachable registry emits an explicit row with exists: null, registry_status: "unavailable" and the error, and a run where no registry answered fails and bills nothing.
2. A near name was presented as a match. Registry name search is not relevance-ranked. Live: querying GLEIF for Siemens AG returns ten records with zero exact matches, led by Siemens Energy AG (CH) - a different, spun-off company - at confidence 0.667. GLEIF registers the parent as Siemens Aktiengesellschaft, which is not on page one at all. With maxPerRegistry: 1 that wrong company was the answer. Every row now carries name_match_grade and verification_outcome, and only an exact normalized name match is ever called verified_exact_name_match.
3. A negative was over-read. US-DE is the Delaware business licence list, not the Delaware corporate registry. Danone US, Inc. is a Delaware entity - Colorado publishes jurisdictonofformation: "DE" for it - and US-DE returned exists: false. Every row now carries registry_scope and registry_false_means, spelling out what a negative on that particular registry does and does not prove.

exists is tri-state: true matched, false the registry answered and nothing matched, null the registry was not checked. Never read null as a negative.

Three status defects fixed at the same time

Registry	What was wrong	Now
US-CO	Registered Agent Resigned (663 live entities) matched the word "registered" and was reported active.	unknown, with the registry's own words in status_raw.
US-DE	Status was hard-coded to Licensed (so, active) while the licence expiry date was fetched and thrown away. An expired Delaware licence read as active.	Status is computed from current_license_valid_to; an expired licence is inactive.
US-PA	typeofbusinessregistration was mapped to status. All 18 distinct live values are entity types ("Domestic Limited Liability Company"), so every one of ~4.5M rows produced unknown from a column that never held a status.	PA publishes no status: status_basis says not_published_by_registry, and the value is surfaced in entity_type where it belongs.
US-NY / US-OR	Active was hard-coded.	Still active - the datasets are scoped to active registrations - but status_basis now reads implied_by_dataset_scope, not fetched.

status_raw carries the registry's own string on every row. The normalizer collapses 15 distinct Colorado statuses and 26 Connecticut ones into four buckets, and status_raw is what was actually published.

Who it's for

• KYB / AML onboarding and vendor-risk teams grounding a company claim in an official registry record before contracting or paying.
• Sales and RevOps enriching and de-duplicating accounts with a registry identifier.
• Fintech and marketplace compliance confirming a merchant entity exists and is in good standing in its stated jurisdiction.
• AI agents that need a chainable verification tool which can say "not checked" rather than guessing.

Registries covered

Code	Registry	What a negative means here
FR-INSEE	French national register (INSEE SIRENE + RNE)	No French entity matched. Says nothing about entities outside France.
GLEIF	Global LEI register	No LEI matched. Most companies worldwide have no LEI, so this is not evidence the company does not exist.
US-CO	Colorado SoS business entities (all statuses)	No Colorado-registered entity matched.
US-CT	Connecticut SoS registrations (all statuses)	No Connecticut registration matched.
US-NY	New York DOS, "Active Corporations: Beginning 1800"	No active New York corporation matched. A dissolved NY entity also returns false.
US-OR	Oregon SoS, "Active Businesses - ALL"	No active Oregon registration matched.
US-PA	Pennsylvania DOS, current registered businesses	No current Pennsylvania registration matched.
US-DE	Delaware business licence list	No Delaware licence matched. Not evidence the company is not a Delaware-incorporated entity - the Delaware Division of Corporations publishes no open bulk dataset.

Iowa was removed in July 2026: data.iowa.gov migrated off Socrata and every Iowa request returned the portal's HTML 404 page.

Example input

{ "companyName": "Danone", "maxPerRegistry": 5 }

Restrict the fan-out with country (FR or US; GLEIF is global and always included) or with an explicit registries list. maxPerRegistry caps candidates per registry, ranked by name-match confidence - but read name_match_grade rather than assuming the top row is the company you meant.

{ "companyName": "Acme Holdings LLC", "registries": ["US-CO", "US-NY", "GLEIF"] }

Output fields (53)

Verification verdict

Field	Meaning
input_name	The company name you queried.
registry	Registry code this row came from.
verification_outcome	verified_exact_name_match, candidate_match_review_required, no_match_on_registry, or registry_unavailable_not_checked.
outcome_reason	Plain-language explanation, including what a negative does and does not prove here.
exists	Tri-state: true matched, false checked and no match, null not checked.
name_match_grade	exact, strong, partial, weak, none. Only exact is a name verification.
match_confidence	0-1 token similarity after legal-form tokens are normalized away. null when not checked.
matched_name	The registry's record name.
entity_id	Registry identifier: SIREN, LEI, entity id, filing number, licence number.

Status

Field	Meaning
status	Normalized active, inactive, dissolved, unknown. null when not checked.
status_raw	The registry's own status string, before normalization.
status_basis	fetched, implied_by_dataset_scope, or not_published_by_registry.

Per-registry outcome contract (on every row)

Field	Meaning
registry_status	ok or unavailable for the registry this row came from.
registry_error	Why it did not answer, when unavailable.
registry_scope	What this registry actually covers.
registry_false_means	What exists: false does and does not prove here.
verification_complete	true only when every requested registry answered and passed its drift assertions.
registries_requested	Every registry this run attempted.
registries_checked	The ones that answered. The result is valid only against these.
registries_unavailable	The ones that did not. null on a complete run.
registry_errors	Per-registry failure detail.
registries_expected_count, registries_checked_count	Coverage of this run.
fr_insee_status, gleif_status, us_co_status, us_ct_status, us_ny_status, us_or_status, us_pa_status, us_de_status	ok / unavailable / not_requested per registry.

The registry record

Field	Meaning
jurisdiction	Registry jurisdiction, e.g. US-CO, FR, GLOBAL.
entity_type	Legal entity type as published.
formation_date	Formation or first-registration date, YYYY-MM-DD.
jurisdiction_of_formation	Where the entity was originally formed - often the key KYB fact.
address	Principal or legal address.
mailing_address	Mailing or agent address, where published separately.
registered_agent	Registered agent or entity of record.
officers	Directors, officers or agent name, where published.
county	County of record (New York, Pennsylvania).
legal_form, registration_authority, entity_category, other_names	GLEIF LEI attributes.
trade_name, license_category, license_valid_from, license_valid_to	Delaware business-licence detail.
activity_code, activity_section, employee_range	INSEE activity and size data for French entities. The INSEE API publishes no human-readable activity label, only the NAF code and section letter.
source_url	Link to the registry record or search page.
retrieved_at	ISO 8601 timestamp of the query.

Everything in the last block except address, officers, jurisdiction, source_url and retrieved_at was already being fetched on every run since v1.0 and thrown away.

Sparse columns, with verified populating inputs

A dead-column audit over 22 live rows from the prefill (2026-08-01) found the Delaware licence block empty, because Danone holds no Delaware business licence. Each sparse column has an input recorded here so it can be re-verified:

Column	Input that populates it (verified live 2026-08-01)
trade_name, license_category, license_valid_from, license_valid_to	{"companyName":"Verizon","registries":["US-DE"]} returns VERIZON SERVICES CORP, category GENERAL SERVICES, valid 2026-01-01 to 2026-12-31
registry_error, registries_unavailable, registry_errors	any run where a registry does not answer, e.g. {"companyName":"Danone","registries":["US-CO","US-XX"]}
legal_form, registration_authority, other_names, entity_category	GLEIF rows only: {"companyName":"Danone","registries":["GLEIF"]}
activity_code, activity_section, employee_range	FR-INSEE rows only: {"companyName":"Danone","country":"FR"}
mailing_address	Colorado and Connecticut rows
county	New York and Pennsylvania rows

Live drift assertions

The dangerous failure here is not an outage. It is a registry that answers HTTP 200 with rows that were never actually filtered - arbitrary companies that read as matches. Every run asserts, before any billable row:

• the name filter was really applied: every row a Socrata registry returns must contain the queried substring in its name column. One that does not fails that registry, which is then reported as unavailable rather than letting unfiltered rows through as matches.
• the dataset was not replaced: at least half the expected columns must still be present. An unknown Socrata column simply comes back absent, which is indistinguishable from an empty value.
• fuzzy results are related to the query: FR-INSEE and GLEIF run their own search, so containment cannot be asserted (querying Danone legitimately returns OLIVIA DANON), but at least one returned record must share a name token with the query.
• a negative control: the nonsense name Qzxjvbmpwl Kdfhrtynugs Holdings is queried once per run and must match nothing. This is what catches a filter that has started matching everything.
• not one registry answered fails the run and bills nothing.

Use as an MCP tool

Callable by AI agents (Claude, Cursor, etc.) via mcp.apify.com. An agent can verify a company and read back verification_outcome, name_match_grade, verification_complete and registry_false_means, so it can distinguish "verified", "candidate", "checked and absent" and "not checked" without guessing.

Pricing

Pay-per-result: one dataset record per registry match, per no-match verdict, or per unavailable registry. A run where no registry answered emits nothing and bills nothing beyond the fractional actor start.

FAQ

Does a false mean the company does not exist? No, and the row tells you what it does mean. Read registry_false_means. Most companies have no LEI, and most Delaware corporations hold no Delaware business licence.

Why does an exact-looking name come back as a candidate? Because registry name search is not relevance-ranked. Siemens AG on GLEIF returns Siemens Energy AG first. Only a match that is exact after legal-form normalization is graded exact.

What if a registry is down? You get a row with exists: null, registry_status: "unavailable" and the error, and verification_complete is false on every row of that run. Nothing is reported as "not found" that was not actually checked.

Do I need an API key? No. All eight sources are keyless public registries.

Related actors

• Sanctions Screening List Change Monitor - OFAC / BIS / State denied-party screening for the entity you just verified.
• Medicaid Exclusion Screener - federal and state healthcare exclusion screening for clinician and vendor checks.
• GLEIF Ownership Graph - parent and subsidiary structure for an entity with an LEI.