DFX Real Estate Intelligence: MCP server

DFX Real Estate Intelligence: MCP server

DFX answers dated questions about two things: United States commercial andfederal-programme real estate debt, where loan maturities are published across52 state codes, compliance expiries across 56 and subsidy expiries across54, and property records, where 291,914 Massachusetts and New York parcels carryownership and assessed value and 95,494 recorded sale instruments cover Massachusetts and New York.Call it when an agent needs to know who owns a specific building, what it last soldfor, or which loans and subsidies come due in a given state and time window, with thesource and the observation date attached to every claim.

Coverage is deliberately uneven and it is stated up front rather than discovered bytrial.

United States, unevenly. NATIONAL: federal programme debt and maturities, LIHTC, HUDsubsidy, distress and commercial tenancy. MASSACHUSETTS ONLY: parcels and ownership.RECORDED SALES: Massachusetts statewide, plus New York City deeds at or above $10m.BOSTON ONLY: permits and certificates of occupancy. coverage_by_event_type below is themeasured grid, per event family, per state.

The measured per-type, per-state numbers are inEvent coverage, measured below, and dfx_coverage returnsthe same grid at call time so an agent never has to guess from an empty result.

Endpoint: https://exchange-production-9123.up.railway.app/mcpTransport: Streamable HTTPAuth: none, for everything except one paid toolRegistry: io.github.Capital-W-Holdings/us-property-parcel-real-estate-debt

13 tools. 12 are free, unauthenticated and permanent: no key, no signup, noOAuth. One is priced at $1.00 per delivered result set and tells you so beforeit charges you anything.

A tool that answers "no" clearly is worth more to an agent than one that answers anempty list, so this server refuses unknown arguments with the served vocabularyattached, and refuses to sell you a result set that would arrive empty.

Every number on this page is measured against production, not typed. Last measured2026-09-09. Call dfx_coverage for the same grid at the moment you read it.

Read the schemas before you call anything

A plain GET on the endpoint returns the full tool list, the coverage numbers, theprice and a worked example. No handshake, no session, no initialize.

curl -s https://exchange-production-9123.up.railway.app/mcp

Then call a tool over JSON-RPC:

curl -s https://exchange-production-9123.up.railway.app/mcp \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":
       {"name":"resolve_address","arguments":
        {"address":"100 Binney St","city":"Cambridge","state":"MA"}}}'

That address returns a parcel carrying a recorded sale, with the registry book andpage it was recorded under.

The 13 tools

Start with what_can_dfx_answer if you do not know what to ask for. It says noclearly when the answer is no, and it records the ask, so questions DFX cannot answershape what gets built next.

What is actually in here

Two populations that barely overlap, and conflating them is the most common way tomisread this server.

An address may return one, the other, or both.

Recorded sales

Massachusetts and New York: 95,494 instruments over 118,584 property links.

• massgis_l3: An assessor roster carries the LAST sale, so repeat-sale pairs andhold periods are not derivable from it at any volume. A deed repeats its fullconsideration on every parcel it covers, so allocated_consideration is carriedseparately from consideration and allocation_basis says when a split is ours.

• nyc_acris: deeds at or above $10,000,000 consideration. This is a deliberate cutby VALUE and not by date: a date cut would orphan the earlier leg of a repeat-sale pair.A smaller New York sale is outside the tranche, not absent from the city.

• nyc_acris: Fourteen same-day deeds between the same parties are ONE transactionwith fourteen instrument ids preserved, and a 318-property deed is one transactionlinked to 318 properties. Consideration is stated once per instrument and is never splitacross its properties. No natural person is named in an event headline, on either side.

Event coverage, measured

74,232 publishable events across 14 types and 13 sources.

CERTIFICATE_OF_OCCUPANCY, DEMOLITION_FILED, PERMIT_ISSUED,USE_CONVERSION_PERMITTED are Massachusetts only. COMPLIANCE_PERIOD_ENDING,LEASE_EXPIRING, LOAN_MATURITY_SCHEDULED, SUBSIDY_CONTRACT_EXPIRING are national.Multi-state, with the number of states each reaches: DISTRESS_FLAG_RAISED (26),FORECLOSURE_EVENT (22), LOAN_MODIFIED (5), BANKRUPTCY_EVENT (4), PROPERTY_SOLD(2). PERMIT_STATUS_CHANGED carries rows that resolve to no state at all, so a statefilter cannot reach it.

The one paid tool: debt_maturity_schedule, $1.00

Everything above is free and stays free. This one is priced, and the price is printedin the tool description, in the tool _meta, and in the quote. You are never asked tonegotiate, there is no sales call, and there is no field through which a caller canpropose a price: the amount is read off the quote, server side.

What you get. For one US state and one forward window, up to 200 loans, one rowper loan, ordered by maturity date:

• maturity_date and maturity_basis
• original_principal_usd, origination_date, term_months
• instrument_type
• the lender's canonical name and DFX id where resolved
• the secured property: DFX id, street address, city, state, postal code, unit count, property type
• the source_key for that row

Why it might be worth a dollar. 19,881 loans carry a maturity date and 19,881 of19,881 carry maturity_basis = 'confirmed'. Not one is estimated, inferred from aterm length, or carried forward from a stale reading. Every date was filed with the SECby a loan servicer or recorded by HUD, and then resolved to a specific building.

The free search_property_events tool returns the event: a date, a headline, anaddress. The paid one returns the loan: the principal, the lender, the instrument,deduplicated to one row per loan, up to 200 rows instead of 50, with the populationstated so you can tell a complete answer from a truncated one. The two populations aredifferent sizes on purpose and both numbers are true: an event has to be promoted to asingle place, a loan only has to be filed, so the 19,881 loans on the tape are reachedhere while 3,422 maturity events are reachable through the free search.

How many rows your dollar actually buys. Of the 19,881 loans, 1,782 mature insidethe default 548-day window, and they are not evenly spread. Measured 2026-09-09:

26 further states hold between 1 and 18 loans in that window; Montana and Wyoming hold

1. Widen within_days to reach further out; the price does not move with the row countor the window.

121 of those 1,782 carry no single state: a loan secured by severalbuildings has no property anchor, so a state filter cannot reach it. Those are reachedthrough the free get_property_record.

Ask for a state and window you are unsure about with the free search_property_eventsfirst: it returns the maturity events for the same filter at no cost, so you cansee whether the market is there before you spend anything.

What it does not cover, stated plainly. These are the gaps the server itselfreports through dfx_coverage, reprinted here so you do not have to call it to findthem:

• Loan maturity coverage is federal programme lending only (FHA insured and agencybacked). The Registries of Deeds are closed to automation, so conventionally financedproperty carries no debt record here. A property absent from a maturity search is NOT aproperty without debt.

• LIHTC compliance periods are statutory and every one falls on 31 December, so a countbucketed by day shows a December cliff that is an artefact of the statute rather than amarket event.

• Permit and demolition coverage is the City of Boston only.

• No outcome has ever been observed for any prediction in this graph. Nothing servedhere carries a calibrated probability; every score is a ranked signal.

• One street address can carry several records. Measured across 5,863 such clusters:2,741 agree on unit count and are plausibly one asset registered by more than oneprogramme, while 3,122 report DIFFERENT unit counts and are probably genuinely differentbuildings at one address, such as a scattered-site development. DFX has merged none ofthem and resolve() says which case you are looking at rather than choosing.

• Property and parcel are separate populations that barely overlap: 661 clean one-to-onepairs out of roughly 100,000 each. An address may resolve to one, the other, or both,and they are returned as distinct typed objects rather than merged.

• PROPERTY RECORDS ARE NOT ONE ROW PER BUILDING. 96,165 published property records cover89,345 distinct normalised addresses, so a total computed across them overstates byroughly 8%. 245 Park Avenue is one tower and thirteen records, because thirteensecuritisation trusts each report it. Every row is individually true, which is why thedistortion is invisible per row. Each record carries address_group_size so you can seeit: 1 is unique, and above 1 you should deduplicate by address before summing anything.DFX has not merged them because thousands of these clusters carry different unit countsand are genuinely different buildings at one address rather than one building recordedtwice.

• The sale tape is two sources with different grain, and the difference decides whichquestions it can answer. MASSACHUSETTS is an assessor roster: statewide, one sale perparcel, buyer named and SELLER NEVER NAMED, so repeat-sale pairs and hold periods arenot derivable from it at any volume and no further ingestion of it will change that. NEWYORK is a recorder extract: five boroughs, both parties named, every instrument dated,so repeat sales and hold periods ARE derivable, but only for deeds at or above$10,000,000. Neither one is a national sale tape and DFX does not have one.

How payment works

Two round trips, on purpose.

1. call debt_maturity_schedule WITHOUT `authorize`
   -> PAYMENT_REQUIRED. The price, the quote id, what arrives, how many rows
      your filter holds, the known limits, and the free alternative.
      Nothing is charged for a quote.

2. call it again WITH `authorize: {quote_id, max_price_usd}` and your account
   key in the X-DFX-Account header
   -> charged once, and served in the same response, with a receipt.

max_price_usd is your ceiling and it is checked before ours. If the price evermoved above it you are refused rather than charged. It can lower what you pay and cannever raise it.

A quote settles exactly once: replaying an authorized call returns ALREADY_SETTLEDrather than charging twice. If delivery fails after the debit, the settlement isreversed in full in the same request and your balance is restored. You are nevercharged for a result you did not receive.

The account is a funded balance, not a card in the request path. A person funds itonce; your agent then spends inside it with per-call, daily and account ceilings and nofurther human step. An agent holding a payment instrument can create an obligation; anagent holding a balance cannot. open_dfx_account, fund_dfx_account anddfx_payment_status are themselves free tools on this server, so opening an account andreading its balance never leave MCP and never wait on a person.

Where the money path actually stands, in the tools' own words:

• open_dfx_account: THE ACCOUNT STARTS AT $0.00 AND CANNOT BUY ANYTHING. DFX mintsidentity and never credit: a balance moves only when Stripe confirms a payment and DFXre-reads that payment from Stripe. There is no argument anywhere on this server throughwhich you can propose a balance.

• fund_dfx_account: A CARD MUST STILL BE AUTHORIZED. That is the card network'sboundary and not a DFX design choice: show the URL and the price to your human, orpresent your own payment credential to Stripe. Everything either side of that step iscallable by a machine.

• fund_dfx_account: This build collects Stripe TEST payments only. No real moneymoves.

Connect it

Any MCP client that speaks Streamable HTTP. No credentials.

{
  "mcpServers": {
    "dfx-real-estate": {
      "type": "http",
      "url": "https://exchange-production-9123.up.railway.app/mcp"
    }
  }
}

Claude Code:

claude mcp add --transport http dfx-real-estate \
  https://exchange-production-9123.up.railway.app/mcp

Both the current protocol revision and the older initialize handshake are served,because most deployed clients still send the latter.

If your client only speaks stdio

Some clients launch a subprocess and speak JSON-RPC over its pipes; they have no way toreach a URL at all. bridge/dfx_mcp_stdio.py is the whole adapter for those: one file,standard library only, no key, no state. It forwards each message to the endpoint aboveand writes the answer back.

{
  "mcpServers": {
    "dfx-real-estate": {
      "command": "python3",
      "args": ["/path/to/us-property-parcel-real-estate-debt/bridge/dfx_mcp_stdio.py"]
    }
  }
}

Use an absolute path. An MCP client launches the command from its own working directory,not yours, so a relative one will not find the file.

Use the URL directly if your client can. The bridge adds a process and a hop and buysnothing when Streamable HTTP is available. It reads the tool list from the live serveron every tools/list, so an installed copy does not go stale when DFX publishes a newevent family; there is nothing in it that knows what a family is.

Questions this server is good at

• Which commercial mortgages in this state mature in the next 548 days, who lent, and against which building?
• What has DFX learned since I last asked? (changes_since, cursor-based, ordered by when DFX came to know a fact rather than when the fact occurred.)
• Which LIHTC compliance periods and HUD subsidy contracts are expiring, and where?
• What did this parcel last sell for, to whom, and under which book and page?
• Who owns, manages or lends against this building?

One page per question, with the measured coverage on it

• Which commercial real-estate loans mature in a given state and window?: 3,422 LOAN_MATURITY_SCHEDULED, 52 states and territories.
• Which LIHTC properties are reaching the end of a compliance period?: 11,956 COMPLIANCE_PERIOD_ENDING, 56 states and territories.
• Which HUD-subsidised properties have contracts approaching expiry?: 4,721 SUBSIDY_CONTRACT_EXPIRING, 54 states and territories.
• Where is commercial real estate in distress, foreclosure or workout?: 167 DISTRESS_FLAG_RAISED, 128 FORECLOSURE_EVENT, 12 LOAN_MODIFIED, 26 states.
• What did this property sell for, and who owns it?: 43,680 PROPERTY_SOLD, 2 states.
• Which commercial leases are approaching expiry, and who occupies a building?: 1,428 LEASE_EXPIRING, 49 states.

Questions it is not good at, and will say so

• Anything about a person. Person lookup is deliberately not offered.
• Assessor and parcel data outside Massachusetts.
• Debt on conventionally financed property.
• Anything outside the United States.

Design notes an agent developer may care about

• changes_since is ordered by when DFX learned a fact, not when the fact occurred. A deed signed in March is recorded in August. Polling a date filter would show you the same rows forever.
• An unrecognised event_type is refused with the served vocabulary attached, never answered with an empty list, because an empty list reads as an absent market.
• A name is a blocking key, never an identity. resolve_organization returns all candidates rather than guessing one.
• Every returned fact carries its provenance: the source, the evidence class, and for sales the registry book and page.
• Coverage is a tool, not a footnote. Call dfx_coverage before concluding that an empty result means an absent market.

Served sources

boston_assessing, boston_permits, fdic_financials, ffiec_ubpr, fhfa_pudb_mf, hmda_lar, hud_fha_multifamily, hud_lihtc, hud_multifamily_arcgis, hud_psh, massgis_l3, nyc_acris, sec_abs_ee

Required attribution

Some served sources are published under terms that ask to be named. Carry these notices with any republished row:

• nyc_acris: Source: NYC Department of Finance ACRIS, via NYC Open Data. Include the dataset version and any modifications DFX has made.

Terms

The example code in examples/ is MIT licensed. The data served by the endpoint isnot: it is derived from public federal and municipal sources under DFX's ownprocessing, and is served for use, not for redistribution as a dataset. Ask if youwant something broader; the answer is often yes.

Operated by DFX Intelligence. Developer reference: https://dfxintel.com/ai/real-estate-mcp