Cities

Ask the city.
Check the answer.

CityTwin is a model of a city built from the documents and data it already owns: master plans, mobility plans, budgets, indicators, open geodata. Ask it a question in plain language and it answers from that model, naming the documents it used and saying plainly what the data cannot support.

Worked model · Brașov, Romania · assembled from public records

CityTwinSão Paulo · 23.5505° S
Reading the model · documents, open data, cadastre
“Where can 400 homes be permitted without new water mains?”
Three parcels. Santana, north edge.
Master plan 2019 Water mains 2024 Sources named

Worked model

What a city looks like once it has been modelled properly.

The figures below come from a working model of Brașov, Romania, assembled entirely from public documents and open data about the city. It is not an official municipal system. It is what the method produces when it is pointed at a city that has never been modelled, and the same method applies to any city anywhere.

15neighbourhoods, with their real administrative boundaries
73.7%impervious surface in the historic centre, the highest of the fifteen, against 3% green
11.1 °Cspread in land surface temperature between neighbourhoods on the same summer day
65.3 hasealed in three years, computed on Copernicus imagery, 2018 to 2021

Copernicus imperviousness · 10 m / px · 2018 to 2021 · model assembled from public records

The gap

Four kinds of tool sit around this problem. None answers the question.

GIS platformsDraw any layer you like. They will not tell you what the layers mean together, or which document backs the number.
Document portalsStore the master plan, the mobility plan and the budget as files. Nobody reads three hundred pages before a council meeting.
BI dashboardsReport last quarter, in tables, to people who need to decide where a building goes.
General AI assistantsFluent about any city on earth, accountable for none of them, with no model of yours underneath.
CityTwinThe city's own record, modelled, so a question returns an answer, on a map, with the source next to it.

Sources first, conclusions second

A plausible answer is not an answer.

Generative models produce fluent text whether or not the evidence supports it. CityTwin is built the other way round: the answer is assembled from the sources inside the city model, so it can be checked before it is used in a meeting, a public document or a funding file.

What it computes

Method written next to every number.

01 Sealed surface

Measured from orbit

Imperviousness density at 10 metre resolution, averaged per neighbourhood, from the European land monitoring service. A measurement, not a land-use assumption.

02 Surface heat

The spread inside one city

Land surface temperature from thermal satellite imagery, compared between neighbourhoods on the same day. Heat islands stop being anecdote and become a number with a date.

03 Walking access

Routed, not drawn

Catchments computed along the street network with its real connectivity. A river with no bridge is a wall, and the model treats it as one.

04 Triage indices

Scored 0 to 100, decomposable

Planting priority from mineral surface against mapped green space; charger deficit from built surface against existing points; solar potential from built surface and annual irradiance.

05 Air quality

Daily series, history kept

Particulate readings ingested automatically and held as a series, so a threshold approach is visible as a trend rather than discovered as an incident.

06 Threshold alerts

Stated method, no black box

Alerts fire against published health and regulatory limits, with the threshold and the method printed beside the alert. On short history it says so, instead of inventing a trend.

Copernicus land monitoring Thermal satellite imagery OpenStreetMap geometry Municipal master and mobility plans Budget and indicator records Air quality sensor networks

How it works

From scattered documents to a decision you can defend.

01 CollectPublic documents, open data and geographic sources, including the urban geometry already mapped in the open. Connectors, not a migration project.
02 ModelThe city becomes zones, projects, problems, services and indicators, each carrying its source and its confidence.
03 AskThe assistant answers from the sources in the model, with assumptions and gaps declared explicitly rather than quietly filled.
04 DecideThe answer appears spatially, scenarios are simulated and compared, and the result exports as a report for the meeting or the funding file.

Data control

The city's data stays the city's.

Every administration knows the fear when it signs with a software vendor: that it will never get its own data back. So the exits were built before the features.

01

Open formats, published schema

The model exports as open JSON against a public schema, and its geographic layers as standard GeoJSON. No proprietary format holds you in place.

02

Straight into your GIS

The model is published as a standard geographic web service. A planner adds the layers to their existing GIS like any other source: no manual exports, no files by email.

03

Hosted where your rules require

The model and the AI processing behind it run in the region you answer to, or on your own infrastructure. Data protection is built in from the design rather than bolted on. No advertising cookies, no profiling.

04

Third-party layers stay theirs

Commercial traffic, satellite imagery and terrain tiles remain under their own licences, and the model says which is which.

Book a briefing

Bring us a city and its documents.

A city with a master plan, a strategy and a budget already has enough to be modelled. Tell us which one, and we come back with what the model would hold, what it would answer, and what it would still not know.

Evaluation access carries synthetic data only. Operational deployments run on your infrastructure.