Docs

Introduction

Fieldgeist Verify is an address model, trained on real US addresses. It reads free text into USPS-standard fields and matches the address to a parcel.

Fieldgeist is an AI lab building tools for real estate. Verify, its first tool, is an address model: US-LEX-1, a purpose-built neural model trained on real US addresses, that reads free text the way USPS Publication 28 defines an address. Behind it sits a parcel index Fieldgeist builds, so every reading can be matched to a property. The API is how you use both today.

Verify does two jobs:

  • Parse. Send one line of free text, get back the address split into USPS Publication 28 components (house number, directionals, street, suffix, unit, city, state, ZIP), the standardized delivery lines, a confidence score and warnings for anything missing or unclear. The model reads every line, at the edge, inside the request.
  • Look up. Send a parsed address, get back the parcel it belongs to: a stable Fieldgeist ID, the county APN, the county FIPS code and coordinates. When the address could be more than one parcel, or needs a unit number to pick one, the answer says so instead of guessing.
curl https://api.fieldgeist.com/v1/verify/parse \
  -H "Authorization: Bearer $FIELDGEIST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"line": "350 5th ave apt 21b new york ny 10118"}'
{
  "input": "350 5th ave apt 21b new york ny 10118",
  "type": "street",
  "components": {
    "city": "NEW YORK",
    "addressNumber": "350",
    "streetSuffix": "AVE",
    "streetName": "5TH",
    "occupancyType": "APT",
    "occupancyId": "21B",
    "state": "NY",
    "zip5": "10118"
  },
  "lines": ["350 5TH AVE APT 21B", "NEW YORK NY  10118"],
  "warnings": [],
  "confidence": 0.9999020747368791,
  "model": { "name": "us-lex-1", "version": "0.1.0", "backend": "model" }
}

Who it's for

  • Proptech teams that need clean addresses and a reliable link from an address to a property record.
  • AI agent builders whose agents call address and property tools and need typed answers with explicit uncertainty (ambiguous, unit_required) rather than a confident guess.

What it covers

ParseLook upLook up by ID
EndpointPOST /v1/verify/parsePOST /v1/verify/lookupGET /v1/verify/lookup/{fieldgeist_id}
InputOne line of textAddress components (or a parse response)A Fieldgeist ID (adr_…)
OutputComponents, delivery lines, confidence, warningsMatch status, parcel IDs, APN, FIPS, coordinates, candidatesOne parcel and its address
ScopeUS addresses: street, PO box, rural route, military, general deliveryStreet addressesAny ID a lookup returned

The parser standardizes what you send; it does not correct spelling. Near-miss street names are handled by the lookup, which flags them with fuzzy: true and a lower confidence.

Start here

On this page