Your portfolio, built from the public record.
Your licence, filings and permits from the city's record, each cited to its document, on your profile and on every address you have worked on — with your agent at that address, reachable by clients and by their agents.
We turn NYC public records into profiles machines can answer over.
One pipeline. The city records a deed in ACRIS, a permit in DOB; the engine ingests each record and updates every profile and graph it touches. Pages, alerts, feeds — all drawn from those structures. It advances with every filing, with no action from you.
Your verified profile, built from the record.
The profile nobody had to write. Volume, trades, completion, license standing — each figure carries the source it was read from and a last-updated stamp. Nothing is asserted without a public record behind it.
The public summary is already live — findable, indexed, rendered for everyone. Claim it to control it; activation unlocks the full record-grounded read.
Your proven network — who you've actually delivered with
The revealed network reads who you build with, weighted and cited. It surfaces your regular team — the architects, engineers, contractors and owners you have filed with, and how often. Not a social graph — a revealed one.
Every edge is clickable and opens the permits behind it. An edge exists only where two nodes co-appear on the same DOB filing; weight = the count of shared filings.
One record under every portfolio.
All your public records, connected
Deeds, permits, filings, and violations joined on the parcel, the license, and the entity — one graph across ACRIS, DOB, DOF, HPD, and PLUTO. It holds relations no single document contains — who hires whom, who builds with whom — every edge computed from co-appearance on filings and cited to them.
Every claim cited
Every fact renders with its source — the ACRIS document ID, the DOB job number, the parcel's BBL. A claim without a citation does not render. This is the input AI answers measurably favor.
Ask anything. So can the machines.
The engine compiles a plain-language question into a query over the record and returns the answer with citations attached. The same engine is served to AI agents over MCP — a buyer's agent researching your market reads the same cited rows.
The data ecosystem
Per-department city sources, cross-linked into one database.The machines meet your client at the point of decision.
People bring decisions to the models — "Should I…?", "What do I do about…?" — and they act on the answers. When the question is a building, a deal, or whom to hire, your profile is what's there to be answered from.
home buyers already research with AI — a third of Gen Z — asking about affordability, neighborhoods, and property values before contacting anyone.
Career and financial questions are among the most common decisions people bring to the models.
People don't just ask — they act on what the models tell them.
LLM knowledge formation
When a buyer asks an AI about your building, your block, or your market, the answer is assembled from documents. Two measured facts decide whether you're in it: models answer best what their corpus documents — accuracy tracks how many documents reference a fact (Kandpal et al., ICML 2023; 2026 survey) — and AI search favors what it can check: content carrying citations and statistics gains up to ~40% visibility in generated answers (GEO, KDD 2024).
Your profiles are exactly that input: factual, dated, structured, every claim carrying its ACRIS or DOB citation. They exist because the city recorded your work — nothing manufactured.
Entity representations
"Who builds in Williamsburg" isn't answered from whoever wrote the best copy. Search and answer engines maintain entity representations — a node per firm: type, attributes, relationships, a confidence score on every fact — and in the LLM era that structured layer is what grounds model answers (Pan et al., IEEE TKDE 2024). The node assembles from references that are consistent, structured, and checkable.
Your profiles are engineered as that input: one identity per actor, keyed to the license and the deed; every fact traced to a recorded document; event linked to building linked to firm.
We don't guarantee rankings, knowledge panels, or a place in any AI's answer. We build what those systems demonstrably favor: structured, cited pages grounded in the public record.
See the cited page a record becomes.
A licence and its filings, published at their own URL: every claim carrying its government document, with the agent that answers for it.
Open a portfolio →The portfolio — per discipline.
The record holds what the city has filed on you; claiming publishes it on your portfolio, at your discipline's own record depth.
Identity and work verified — license roll plus DOB filings: permits, inspections, sign-offs.
Identity and work verified — license roll plus every filing signed as applicant of record.
Identity and work verified — DOB registration plus every job you filed as the filing representative.
Identity on record (license roll); work verification in progress — deed-matching your closed deals.
Identity on record (license roll); work verification in progress — closing-counsel data.
The record fills every profile as deep as the record goes — every discipline joins at its own depth, and deepens as records open.
These aren't promises. They're rules the system can't break.
Each rule is built into the software itself — if a query would break one, it simply can't run. We can't quietly change our minds later, because there's no setting to change.
The city records everything it does — every sale, every permit, every deed — then buries it where no buyer will ever look. Meanwhile the people it matters to have started asking an AI instead of digging.
Every AI product aimed at real estate so far points inward — automate the paperwork, draft the email. That's efficiency. Urbicana does the opposite: it takes the activity the city already publishes and turns each event into a cited page — fed to Google, the LLMs, and any agent that asks. You don't write it or optimize it. The record does the work.
Your record is already being filed. Claim the portfolio it builds.
$1,500 setup, once, and $149 a month: your portfolio built from the record, cited to every document, with your agent at its address.