NYC ACRIS Deed Transfer Intel - LLC, Cash & Out-of-State Buyer Flags

A NYC deed-transfer intelligence feed built on the city's official ACRIS property records: who is buying, where, and with what money. Every recorded deed is joined across three ACRIS datasets into one row carrying the BBL, sale price, buyer and seller names and full mailing addresses, the parcel's property type, and analytics flags no raw ACRIS dump gives you:

• llc_buyer - the grantee is an entity (LLC / LP / LLP / TRUST / CORP / INC / HOLDINGS / PARTNERS), not a person
• out_of_state_buyer - a grantee mailing address is outside New York
• cash_sale - no mortgage was recorded against the same parcel around the deed date
• non_market_transfer - the deed recorded $0 consideration (intra-family, trust or estate transfer, not a sale)

Runs on the keyless NYC Open Data (Socrata) ACRIS datasets - no login, no API key, no anti-bot games.

What changed in v1.1 (2026-08-01)

The borough filter was measuring the wrong thing. v1.0 selected deeds on Master.recorded_borough, which is the ACRIS recording office code, not the borough the property is in. Measured live over 1,500 consecutive deeds: recorded_borough matches the property's borough on only 19.8% of deeds (297/1,500), and 94% of all ACRIS documents are recorded under borough 1. The practical effect on a Brooklyn run was that the actor could see only 154 of an estimated 3,841 Brooklyn deeds in a 60-day window, and 33 of every 100 rows it returned were not in Brooklyn (15 Manhattan, 4 Bronx, 14 Queens on the 2026-08-01 run).

v1.1 selects on the property borough from ACRIS Legals. On the same input the run is now 100/100 Brooklyn, and 98 of those 100 deeds were recorded under borough 1 - i.e. invisible to v1.0. The old signal is still published, as recorded_borough.

Also in v1.1: cross-run delta memory actually persists (it never did - see FAQ), every row states which sources answered, buyer and seller mailing addresses are emitted instead of being fetched and discarded, condo unit numbers are included, multi-parcel deeds report every parcel, and property-type codes are decoded from the authoritative published table.

Who it is for

• Real-estate agents farming just-sold lists: new owners are the warmest listing / refi / insurance leads in the market, and the buyer's mailing address is now on the row.
• Investors & analysts tracking institutional activity: filter llc_buyer=true + cash_sale=true + non_market_transfer=false to see where entity cash is landing, block by block.
• Title, lending & proptech teams feeding transfer events into their own pipelines - mode: "delta" gives a genuine incremental feed.
• Journalists & researchers studying out-of-state and entity ownership of NYC housing.

Example input

{
  "boroughs": ["3"],
  "docTypes": ["DEED"],
  "lookbackDays": 60,
  "mode": "inventory",
  "maxResults": 100,
  "cashSaleToleranceDays": 45
}

boroughs selects on the property borough (1 Manhattan, 2 Bronx, 3 Brooklyn, 4 Queens, 5 Staten Island); leave it empty for all five. Set mode to "delta" on a schedule to receive only newly recorded deeds each run.

Output fields

Every field below is emitted on every row. null always means not checked - never "checked and negative".

Field	Description
document_id	ACRIS document identifier shared across Master, Legals and Parties.
recorded_datetime	Date the deed was recorded (YYYY-MM-DD).
doc_type	ACRIS document type code (e.g. DEED).
borough	PROPERTY borough code from ACRIS Legals: 1 Manhattan, 2 Bronx, 3 Brooklyn, 4 Queens, 5 Staten Island. This is what the boroughs input selects on. Not the same as recorded_borough.
block	Tax block from Legals.
lot	Tax lot from Legals.
bbl	10-digit Borough-Block-Lot key (joinable to PLUTO and DOF datasets).
street_address	Street number + name from the first Legals record. Does NOT include the unit - use street_address_full for mail.
price_usd	Deed consideration (document_amt) in USD. 0 is what ACRIS publishes for a non-market transfer (see non_market_transfer); null means the field was not published.
seller_names	Grantor (party_type 1) names.
buyer_names	Grantee (party_type 2) names.
buyer_mailing_state	Mailing-address state of the first grantee.
llc_buyer	True when any grantee name matches an entity pattern (LLC, LP, LLP, TRUST, CORP, INC, HOLDINGS, PARTNERS...). NULL when the Parties source did not answer or no grantee was recorded - null means not checked, false means checked and negative.
out_of_state_buyer	True when any grantee mailing state is not NY. NULL when the Parties source did not answer or no grantee state was recorded.
cash_sale	True when no mortgage (MTGE) was recorded against the same BBL within the tolerance window (default +/-45 days). NULL when the BBL is unknown or the mortgage lookup did not answer for that parcel - see mortgage_status. Note 56% of deeds in the default window are $0 non-market transfers, which are cash by definition; combine with non_market_transfer and price_usd.
source_url	ACRIS document view link for the deed.
borough_name	Property borough name decoded from the Legals borough code.
recorded_borough	Master.recorded_borough: the ACRIS recording-office code, NOT the property borough. Measured live 2026-08-01 it matches the property borough on only 19.8% of deeds, so it is reported for reference and never used to select rows.
recorded_borough_name	Name for recorded_borough.
borough_filter_basis	Which borough the boroughs input was applied to. Always "property_legals" from v1.1 onwards.
unit	Condo / apartment unit from Legals. Populated on 31.2% of Legals rows; null when the record carries none.
street_address_full	street_address plus the unit when one is recorded. Use this for mail; street_address alone is undeliverable for a condo.
property_type	ACRIS Legals property_type code (PA, SC, D1, D2, D3, AP, CR...).
property_type_desc	Decoded property_type. Null when ACRIS publishes a code outside the documented set.
parcel_count	Number of Legals (parcel) records on this document. 8.1% of sampled deeds cover more than one parcel. Null when the Legals leg did not answer.
additional_bbls	BBLs of the second and subsequent parcels on a multi-parcel deed. Empty array when single-parcel; null when the Legals leg did not answer.
easement	Legals easement flag. Null when not recorded.
air_rights	Legals air-rights flag. Null when not recorded.
subterranean_rights	Legals subterranean-rights flag. Null when not recorded.
partial_lot	Legals partial-lot indicator (E = entire lot, P = partial).
document_date	Date on the instrument itself, which precedes the recording date.
crfn	City Register File Number.
percent_transferred	Master.percent_trans. 100 for a full transfer. Null when not published.
partial_interest_transfer	True when percent_transferred is below 100. Null when percent_transferred is unknown.
non_market_transfer	True when the recorded consideration is exactly $0 -- typically an intra-family, trust or estate transfer rather than a sale. 56% of deeds in the default window are $0. Null when the amount was not published.
buyer_count	Number of grantee (party_type 2) records. Null when the Parties leg did not answer.
seller_count	Number of grantor (party_type 1) records. Null when the Parties leg did not answer.
buyer_mailing_states	Every distinct grantee mailing state, deduplicated. buyer_mailing_state reports only the first grantee and can disagree with out_of_state_buyer; this field does not.
seller_mailing_states	Every distinct grantor mailing state, deduplicated.
buyer_mailing_address	First grantee address_1. Populated on 100% of sampled deeds.
buyer_mailing_address_2	First grantee address_2. Populated on 8.1% of party records.
buyer_mailing_city	First grantee city. Populated on 100% of sampled deeds.
buyer_mailing_zip	First grantee ZIP. Populated on 99.7% of party records.
buyer_mailing_country	First grantee country code.
buyer_contacts	Every grantee as {name, address_1, address_2, city, state, zip, country} - the full mailing block ACRIS publishes. Null when the Parties leg did not answer.
seller_contacts	Every grantor as {name, address_1, address_2, city, state, zip, country}. The grantor address is often the seller forwarding address. Null when the Parties leg did not answer.
other_party_names	Names recorded with party_type 3 (neither grantor nor grantee).
master_status	"ok" or "partial" - whether every ACRIS Master page was retrieved.
legals_status	"ok" (joined), "no_legals_recorded" (upstream answered, nothing recorded) or "unavailable" (the request failed). "unavailable" means the property fields on this row are unknown, not absent.
parties_status	"ok", "no_parties_recorded" or "unavailable". When "unavailable", llc_buyer and out_of_state_buyer are null rather than false.
mortgage_status	"ok", "partial" (candidate discovery was capped or a batch failed for this parcel), "unavailable" or "not_applicable" (no BBL). cash_sale is null unless this reads "ok".
partial_data	True when any contributing source did not fully answer for this row.
data_through_date	ACRIS good_through_date for the document - how current the source record is.
delta_baseline_status	"ok" (a real cross-run baseline was loaded), "seeded_this_run" (first run for this scope; the window is emitted as the initial inventory) or "not_applicable" (inventory mode).
run_mode	"inventory" or "delta".
window_start	First recording date included, counted back from the dataset maximum.
window_end	Last recording date included = the newest recording in ACRIS Master.

Field population, measured live on the default input (100 rows, 2026-08-01)

Contact and identity fields, because a lead list whose contact column is thin but advertised without qualification is a wrong answer:

Field	Populated
buyer_names	100%
buyer_mailing_address	100%
buyer_mailing_city	100%
buyer_mailing_state	100%
buyer_mailing_zip	100%
buyer_mailing_country	100%
buyer_contacts (full block per grantee)	100%
seller_names / seller_contacts	100%
street_address / street_address_full	100%
bbl, block, lot, property_type	100%
buyer_mailing_address_2	14% - ACRIS publishes a second address line on only a minority of party records
unit	28% - only condo / co-op / unit-level parcels carry one

Zero of the 57 declared fields were empty across the run. ACRIS does not publish phone numbers or email addresses for deed parties, so this actor never claims to: the contactable identity it gives you is a name plus a postal mailing address.

Data sources (official NYC Open Data, all keyless)

• ACRIS Real Property Master (bnx9-e6tj, 17,036,716 rows) - document type, recording date, consideration, CRFN, percent transferred
• ACRIS Real Property Legals (8h5j-fqxa, 22,688,577 rows) - borough / block / lot, street number & name, unit, property type
• ACRIS Real Property Parties (636b-3b5g, 46,456,160 rows) - grantors, grantees, full mailing addresses
• ACRIS Property Types Codes (94g4-w6xz, 48 codes) - the authoritative property_type decode

Live drift assertions

The actor checks the upstream before it produces a single billable row, and fails the run loudly if any of these regress:

• all three ACRIS datasets resolve and still expose every column the join reads
• a negative control (doc_type='ZZZZ_NOT_A_REAL_DOC_TYPE') returns exactly 0 rows - proof the $where clause is being applied
• party_type stays inside its closed vocabulary {1, 2, 3}; an unknown value fails rather than being silently bucketed as neither buyer nor seller
• the property-borough vocabulary stays inside {1..5}
• the newest recording in ACRIS is within 400 days of today (measured lag on 2026-08-01: 32 days)
• at least 50% of scanned documents join to a Legals record (measured: 100%)
• the property-type code table still publishes at least 40 codes (measured: 48)

Every measured value is written to the run log, so tightening a band later is evidence-based rather than a guess.

Use as an MCP tool

Callable by AI agents (Claude, Cursor, etc.) via mcp.apify.com. Field-level output descriptions make it a clean chainable tool for "who bought this / what sold near X" workflows. The <source>_status fields let an agent tell a verified negative from an unanswered question.

Honest notes & limitations

• Recording lag. ACRIS open-data recordings trail real-world closings by several weeks to ~2-3 months. The lookback window is anchored to the latest recorded date in the dataset, never to today, so a 60-day window always returns the freshest available deeds.
• cash_sale is a heuristic, and it is now a gated one. true means no MTGE document was recorded against the same BBL within +/- cashSaleToleranceDays (default 45) of the deed. It is null, never true, when the mortgage lookup could not be completed for that parcel - check mortgage_status. Seller financing and very late mortgage recordings are genuine edge cases.
• $0 deeds are common. 39 of the 100 rows in the default run recorded $0 consideration. Those are intra-family, trust or estate transfers, and they are cash by definition - so cash_sale=true on its own is not an all-cash purchase. Combine with non_market_transfer = false.
• Multi-parcel deeds. bbl is the first parcel; parcel_count and additional_bbls carry the rest. 8.1% of sampled deeds cover more than one parcel.
• Addresses are as recorded. street_address is the Legals street number + name; street_address_full appends the unit. Neither is USPS-normalised.
• Staten Island is thin. 206,662 of 22.7M Legals records, because Richmond County keeps its own index. A borough-5 run can legitimately return very little.
• No phone or email. ACRIS does not publish them for deed parties.

FAQ

How do I get just-sold leads for my farm area? Set boroughs to your borough, mode to delta, and schedule the actor weekly. Each run emits only newly recorded deeds, with buyer names and mailing addresses ready for a CRM.

Did delta mode really not work before? It did not. v1.0 kept its "already seen" memory in the run's default key-value store, which Apify creates fresh for every run - eight consecutive runs on 2026-08-01 each had a different store ID. Every delta run therefore reported itself as the first one and re-emitted (and re-billed) the whole window as new. v1.1 uses a named store, keyed by your borough / document-type / window combination so two differently-filtered schedules cannot overwrite each other, and every row carries delta_baseline_status (ok, seeded_this_run or not_applicable) so you can tell an incremental run from a seeding one.

How do I track institutional / LLC buyers in NYC? Filter llc_buyer = true, optionally with cash_sale = true, non_market_transfer = false and price_usd > 1000000. Grantee names are the LLC names as recorded on the deed, and buyer_contacts gives the LLC's service address.

Why is llc_buyer sometimes null instead of false? Because the actor did not verify it. null means the ACRIS Parties leg did not answer for that document, or no grantee was recorded; false means grantees were read and none looked like an entity. parties_status tells you which.

Why is the sale price $0? ACRIS records the deed consideration. Transfers between family members, into an owner's own LLC, or via estates commonly record at $0. Use non_market_transfer or price_usd > 0.

Does it cover all five boroughs? Yes - selection is on the property borough from ACRIS Legals. Note the caveat on Staten Island above.

Pricing

Pay per result: you are charged only for the deed records actually emitted. A run that fails a drift assertion emits nothing and bills nothing.

Related actors - the NYC investor shelf

Same BBL keying, built to run together:

Actor	What it adds
NYC Distressed Property Leads	Motivated-seller scoring - open HPD violations, tax-lien-sale notices and executed evictions per parcel
NYC Landlord Registry Lead List	Owner / agent contacts rolled up by portfolio size
Parcel Owner Lookup	Address to owner of record for single-parcel checks

Typical play: spot an LLC or cash buyer here, look up that buyer's wider portfolio in the HPD registry, then screen their other parcels for distress.