Your Arosa account
Sign in with Arosa. Conversations, properties, concepts and sources are saved to your own workspace, with history and search across your devices.
Know what a site could become.
Arosa is an AI research partner for development in any US city. Name an address with its city and state, and it reads that jurisdiction's current official zoning sources with citations, models the concepts in 3D, and runs the numbers. In New York City it goes deepest today: the parcel resolved from city records, the envelope computed, the model set among the real neighbours. Every claim keeps its source. Every number keeps its assumptions. And every answer names the coverage it rests on.
Arosa · 120 Broadway, Manhattan
Example · NYC deep coverageRunning
- Any US city
- The jurisdiction is detected from the address. Any US city gets cited research against its own official sources.
- Deepest in NYC
- All five boroughs resolved from current PLUTO parcel records, with computed envelopes and footprint neighbours.
- Cited
- Every regulatory claim carries the source it rests on, with its jurisdiction and retrieval date.
One project state
Change one assumption. Everything downstream re-derives.
Parcel, zoning, envelope, concepts, program, market inputs and pro forma live in one state with a dependency graph. When the property or an input changes, the work that depended on it is marked out of date and re-run, so a plan, a number and a deck never disagree with each other.
Each field records which tool produced it, which run wrote it, and the sources it rests on, with their retrieval dates.
Pick a change to see what goes out of date.
Fourteen skills
An agent, not a wizard.
Ask for any step at any time. The canonical flow still reads in order, and the fourteenth skill runs it end to end: site intake, zoning, entitlement, massing, program, parking, plans, pro forma, recommendation, concepts, deck.
In Arosa today
01
Site intake
An address with its city and state becomes an identified site in its jurisdiction. In New York City, an address, a building name or a BBL becomes a confirmed parcel with its boundary, the neighbours from NYC building footprints, and nearby transit, parks and retail.
02
Zoning engine
The site's own zoning code, read live and cited by section in any city. In New York City, also districts, overlays and special districts from the parcel record, FAR by use against the existing building, and curated rules with their Zoning Resolution sections, read as starting points to verify.
03
Entitlement
The mechanisms that could add floor area or change use, with the approvals and recorded instruments each one needs.
04
Massing study
Concepts drawn in 2D and 3D with computed areas, FAR and height, compared side by side in any city. In New York City they are also placed among the footprint neighbours, with Street View and photorealistic context.
05
Program and unit mix
Market inputs recorded with their sources, and a unit mix by bedroom type sized to the residential area and the efficiency you expect.
09
Financial model
Development budget, revenue, NOI, yield on cost, value, returns and sensitivities, run as code from the inputs you supply, with the arithmetic shown.
Next on the roadmap
06
Parking optimization
Required and achievable parking, layouts and ramps, under each jurisdiction's own rules; the Manhattan Core rules are encoded today.
07
Preliminary plans
Cores, corridors and unit layouts per storey, exported for the architect to continue from.
08
Efficiency optimization
Net-to-gross, unit counts and layout choices tuned against each other.
10
Feasibility recommendation
The concept worth pursuing, with the evidence and the risks laid out.
11
Architectural concepting
Facade and material directions that fit the site and the program.
12
Iterative design
Version history for concepts; revise, branch and compare.
13
Final deliverable
A deck drawn from the project state: site, zoning with citations, scenarios, massing, program, pro forma.
14
Full underwriting and plan
One request, every skill in order, pausing only where your judgment is needed.
How it works
From an address to a defensible answer.
Sources
Built on the official record.
Arosa reads current adopted text rather than a handbook or a model's memory, and shows each source with its jurisdiction and date so you can inspect it. The registry below is the deep New York City stack; in a research-tier city the same discipline applies to that jurisdiction's own zoning code, maps and agency records, retrieved live for the question.
NYC Planning ZoLaParcel discovery, zoning districts and mapped context.
NYC Zoning ResolutionThe current adopted text, cited by section, with the linked official zoning maps.
NYC Department of BuildingsBIS and DOB NOW records for the existing building and its filings.
PLUTO and MapPLUTOVersioned tax-lot data: lot size, frontage, built area, districts, landmark and historic district flags.
NYC building footprintsThe neighbours and the existing building, placed around the modeled lot.
Google Maps PlatformNearby places, the Street View photo facing the lot, and photorealistic 3D tiles under the model.
Cited, not asserted. A regulatory claim without a source is not made.
Starting points, not determinations. Curated tables are read as figures to verify against current text and a survey.
Calculation as code. Areas, capacity, unit counts and the pro forma run as inspectable code; the model explains and chooses.
Coverage
Deep where the records are wired. Cited everywhere else.
Arosa detects the jurisdiction from the address you name: a borough or a BBL means New York City, a city with its state means that jurisdiction. It never runs another city's address against New York records, and never borrows New York bulk rules for a site elsewhere. When the city and state cannot be determined, it asks.
New York City: deepParcel resolution from PLUTO and GeoSearch by address, borough or BBL, with nearby-parcel lookup and the property panel.
Computed envelopeCapacity from PLUTO floor-area fields, plus per-storey envelope geometry for the districts encoded so far, mostly in Lower Manhattan.
Real neighboursSurrounding buildings and the existing building from NYC building footprints, with nearby places, Street View and photorealistic tiles.
Any other US city: researchLive research against that jurisdiction's official zoning code, maps and agency sources, cited by section with retrieval dates, plus saved sources scoped to the city, its state and federal parents.
Concepts and numbers2D and 3D massing from a lot outline you supply or retrieved evidence supports, with computed areas and FAR, then a unit mix and pro forma from sourced market inputs.
Not there yetNo parcel-record adapter, computed bulk envelope or footprint neighbours outside New York City. An answer there says so, and treats district, height and floor-area questions as things to read in the local code, not to infer.
Named, not assumed. Every answer states its jurisdiction and coverage tier before it states a rule.
Gaps stay gaps. An unchecked issue remains unknown rather than cleared, in a research-tier city as in New York.
Depth grows city by city. A parcel adapter, curated rule tables and footprint context are added deliberately per market, then verified against that city's records.
Access
Private by design.
Arosa inside ChatGPT
The New York City parcel, zoning and evidence tools are available through an OAuth-protected MCP endpoint, so deep-coverage research can continue in the tools you already use.
Start with an address
- Open Arosa and name a site with its city and state.
- Ask what could be built as of right, and what would take an approval.
- Compare the concepts worth pursuing, with the numbers and the coverage behind them.
Arosa is open to invited accounts during the pilot.