Urban planning
Zoning impact, parcel rollups, transit accessibility — query millions of parcels without writing a single ST_* function.
- Parcel-level zoning analysis
- 15-min city accessibility scoring
- Permit clustering & trend detection
TerraQuerry is the natural-language interface for your geospatial data. Connect PostGIS, Esri, or raw GeoTIFFs — then ask the questions you've always wanted to ask of your maps.
I found 1,247 parcels intersecting the 100-year flood zone within 2 km of the Cedar River — about 412 acres totaling $890M in assessed value. Highest concentration sits along the river's northern bend (outlined orange).
Most LLMs hallucinate when you ask them spatial questions. TerraQuerry doesn't, because it never guesses — it indexes your data, plans a spatial query, executes it on a real engine, and grounds every answer in the geometry it touched.
Connect PostGIS, ArcGIS, S3, or upload Shapefiles, GeoJSON, GeoTIFF, KML. Schemas, projections, and CRS are auto-detected.
Geometry goes into spatial indices (R-tree, H3). Metadata, attributes, and column semantics are vectorized into a retrieval store the planner can search.
You ask in English. The planner decomposes your question into spatial primitives (intersect, buffer, nearest), runs them, and returns geometry — not just text.
"Parcels in 100-year flood zone within 2km of Cedar River."
SELECT p.* FROM parcels p
JOIN flood_100yr f ON ST_Intersects(p.geom, f.geom)
WHERE ST_DWithin(p.geom, cedar_river, 2000);
1,247 features · GeoJSON + map + cited rows
Each example runs against a sample dataset of 1.4M parcels, 28k flood polygons, and 9 years of NDVI rasters covering the King County region.
I found 1,247 parcels intersecting the 100-year flood zone within 2km of the Cedar River — about 412 acres totaling $890M in assessed value. The highest concentration sits along the river's northern bend (outlined in orange); 38% are residential and 11 are critical facilities.
parcels ⨝ flood_100yr · ST_DWithin(cedar_river, 2km)
1,247 parcels · 412 ac · $890M aev
Zoning impact, parcel rollups, transit accessibility — query millions of parcels without writing a single ST_* function.
When the wildfire perimeter shifts, ask: "Which substations are inside the 6hr forecast cone?" — in seconds, not GIS days.
Air-gapped. SCIF-friendly. On-prem deployment that keeps your tasking, telemetry, and trajectories on your network — never ours.
Optimize routes, depot placement, and last-mile zones with constraint-aware queries. Ask "what if?" instead of waiting for a model run.
The planner understands buffer, intersect, contains, nearest, isochrone, viewshed, and 40+ other primitives.
Mix GeoTIFF, NetCDF, Zarr, COG with vector layers in the same query. NDVI math, zonal stats, hillshade — all callable.
"How did this change between 2019 and now?" Temporal queries are first-class. Snapshot diffing, trend detection, and back-fill.
Reprojects between EPSGs on the fly. Datum shifts, geographic vs projected, geocentric XYZ — sorted, not your problem.
Every result includes the rows, polygons, and rasters it touched. Drill into the geometry behind any number.
Python, JS, REST. Drop in a Jupyter cell, embed in your dashboard, or build agents on top. GeoJSON in, GeoJSON out.
Deploy in your VPC, your data center, or a SCIF. Zero egress. Air-gapped install. Models you control.
GeoJSON, GeoParquet, FlatGeobuf, MVT, COG, STAC. No proprietary lock-in. Pull your data and your indices any time.
Live connectors to the tools your team already runs. No re-tiling, no migration. Read straight from source.
We built TerraQuerry for organizations that can't send their geospatial data over the public internet. On-premise, air-gapped, classified — the same product, on your hardware, with your models.
Private beta is open to GIS, intel, and infrastructure teams. Bring a dataset, leave with a working agent on top of it.
Tell us a bit about you and we'll reach out within 2 business days. Beta seats are batched weekly.
We'll reach out from access@terraquerry.ai with next steps. In the meantime, the docs are open.