IHI Super API Hub is the property-intelligence stack other PropTech vendors wish they could ship - 259 canonical signals per parcel, drawn from 1,068 source-adapter connectors, every cell stamped with an R2 sha256, a source endpoint, and a fetched_at timestamp. There is no AI in the runtime - only deterministic fetch, validate, and serve - because hallucinated data on a six-figure deal is not a feature. If you operate in real estate, lend against it, insure it, or build on top of it, this is the stack you rent instead of building for three years.
Why we don't use AI to decide anything
Every PropTech vendor is selling you "AI." Most of those models hallucinate. They cannot show their work. They ship wrong data that loses real money on real deals, and when you call to complain, support points at a model and shrugs.
IHI is the inversion. Our runtime is mechanical: a value is either fetched from a named source endpoint at a known timestamp and hashed in R2, or it is not in the response. There is no model in the path that decides what a signal "should be." Lead scoring is deterministic Python with fixed coefficients - you can read it, audit it, and reproduce it on your laptop. AI lives in two lanes: drafting documents you review before publish, and monitoring anomalies you act on. Never in the value chain.
Sue us if a signal is wrong - we hand you the R2 sha256, the source endpoint, and the fetched_at timestamp. That's the moat.
The proof pillar
R2 object keyContent-addressed sha256 of the raw bytes we pulled. Same bytes in, same key out.
Source endpoint URLThe exact government or public API we hit. You can re-fetch it.
fetched_at timestampUTC ISO-8601 to the second. Recency is a property of the cell, not a marketing claim.
JSON-schema-locked payloadSchema sha 7ecc34fabf24..., version 3.3.0. Contract changes ship as a version bump and a migration.
What's in the box
259-signal master schema
16 layers, sha-locked to v3.3.0. Geo, physical, legal, market, capital, environment, municipal distress, owner, commercial, business, vision.
1,068 source adapters
Federal, state, county, city, vendor (Tyler Munis, Accela, Socrata, ArcGIS, Kofile, Fidlar). Pre-mapped.
Per-county rate-limit DOs
Fair-share at the edge. We do not get our shared county sources banned.
Vision distress detection
NAIP/USGS ortho-tiles, Blue Tarp + Overgrowth + roof condition classification. Runs nightly per county.
Deterministic lead scoring
108-signal propensity engine plus 8 IHI Secret synergies. No LLM in the hot path.
ProofRail audit ledger
Every consequential event hashed and ledgered. Pass an audit, not a vibe check.
Pricing
Tier
Monthly
Counties
Routes
Trial
$0 (14 days)
1
Property lookup
Bronze
$499
5
+ Signals list
Silver
$1,499
25
+ Lead scoring + Lead packets
Gold
$4,999
100
+ Vision/ortho + Skip-trace + Dedicated D1
Platinum
Custom
Unlimited
Full API + Durable Objects + Custom domain
Frequently asked
How is this different from CoreLogic or ATTOM?
They aggregate licensed data they buy from third parties and re-sell. You get a black box. We pull from public sources, hash every byte, and hand you the chain of custody. Plus our schema lock means contract changes are versioned, not surprise breakages.
What happens if your data is wrong?
You get the R2 sha256, the source endpoint URL, the fetched_at timestamp, and the schema version. You can re-fetch the exact source yourself. If we ingested wrong, we file a correction PR against the connector and ship a v-bump within 24 hours. If you sue, the trail is in your hands.
Can I audit the source of any specific signal?
Yes. Every signal on every property comes with its R2 key. curl the proof endpoint and you get back the raw bytes we processed. No black boxes.
Do you use AI to fill in missing data?
No. A gap is a gap. Hallucinating numbers into a deal pipeline is how PropTech loses lender trust. We ship null with provenance, not fabricated values.
How fast can I onboard a new county?
If the county is in our 1,068-adapter library, minutes. If it is a new vendor pattern, we add it within a week using the connector template. New-jurisdiction onboarding is a real product line, not bespoke services.
Is this real-time?
Sub-100ms property lookups (KV-cached at the edge). Underlying public sources refresh on their own cadence - county assessor data is daily-to-quarterly. Your fetched_at timestamp tells you exactly how fresh each cell is.
Where does the data come from?
Federal (Census, HUD, FEMA, EPA, IRS NFTL, SEC EDGAR), state (DMV, licensing, secretary-of-state), county (assessor, recorder, tax, sheriff, permits, code, GIS), city (Socrata portals), and vendor platforms (Tyler, Accela, ArcGIS, Kofile, Fidlar, ePropertyPlus). 1,068 mapped adapters and growing.
What stops me from doing this myself?
Three years and a half-million dollars of source-adapter work. We already did it. Your alternative is paying a fractional data engineer to chase 3,222 counties one at a time and rebuild ProofRail.
Ad-variant tests
Property data with a receipt.
Sue us. We saved you the sha256.
Built mechanical, not magical.
Start a 14-day trial - 1 county, free
We turn on a Bronze sandbox in 15 minutes. You query one county. If the data is not provable to your auditor's satisfaction, you walk.