Your geodata files, on the cloud, readable from anywhere.

Why cloud-native

Your data, readable in place by any modern tool.

Classic GIS files have to be downloaded whole, or put behind a map server someone must run. Cloud-native formats are organised so that each tool reads only the part it needs, straight from where the file is stored. Faster, cheaper, and nothing to maintain. Geobox makes them for you, with the right settings.

Already have cloud-native files online? Link them: Geobox reads them where they are, without copying them.

How it works

From the files you have to data any tool reads.

1

Drop your files

Shapefile, GeoPackage, File geodatabase, GeoJSON, CSV, KML, GeoTIFF, JPEG 2000… or a link to a file, a map service (WFS, ArcGIS) or a STAC catalog. Geobox stores each one as GeoParquet or Cloud-Optimized GeoTIFF, sorted and compressed for reading online, and keeps your original.

Files that are already cloud-native stay where they are: link them, Geobox reads them in place. Ready-made open data is there too, in our catalogs: OpenStreetMap, the world's administrative boundaries, US Census boundaries and numbers (ACS), FEMA flood zones and National Risk Index, protected areas (PAD-US), US addresses and place names, Natural Earth, Geoconnex water features, MassGIS, Corine land cover. Add them to your box as links, or query them where they are.

2

One link for every tool

Each dataset has its address. Paste it in DuckDB and query it in three lines, add it to QGIS, read it from Python: only the parts a query needs travel, never the whole file.

Private datasets stay private: the address goes through your workspace's own key. Or download any of them as GeoPackage, Shapefile, GeoJSON, CSV, KML or FlatGeobuf.

5

And a map, if you want one

Every dataset opens on a map in your browser, millions of features included, and maps of several layers can be public: each layer says where its data comes from and under which licence. Styles, classes and the rest are in Help.

Acknowledgements

Built on open source.

Geobox stands on open source work by people and communities who share it freely. Thank you to:

• geoparquet-io, by Chris Holmes and contributors: every file becomes an optimised GeoParquet through it.
• tylertoo, by Nissim Lebovits and contributors: GeoParquet to PMTiles, the map tiles of every vector dataset.
• GeoParquet, the open specification, and the OGC working group behind it; GeoArrow, by Dewey Dunnington and contributors.
• DuckDB, its spatial extension and DuckDB-Wasm, which read and analyse the data on the converter and in your browser.
• H3, the hexagon grid started at Uber, and its open source community; the H3 extension for DuckDB by Isaac Brodsky: the hexagons of the zoomed-out maps.
• PMTiles, by Brandon Liu and Protomaps.
• GDAL and PROJ, which read almost every geographic format and every coordinate system there is.
• MapLibre GL JS, which draws the maps, and geotiff.js and proj4js for imagery in the browser.
• async-geotiff and obstore, by Kyle Barron and Development Seed.
• Apache Arrow and Parquet, the ground all of this is built on.
• Python and NumPy, which the converters run on.
• vector-tile and pbf, by Mapbox: the thumbnails in your box, drawn from the map tiles.
• Inter Tight, by Rasmus Andersson, through Fontsource: the font of the titles.
• esbuild, by Evan Wallace, and uv, by Astral, which build and package Geobox.
• STAC, the catalog specification your published datasets follow.
• Portolan and its community: the conventions the Geomermaids catalogs and your public catalog follow.
• The Sentinel-2 images of the Copernicus programme (ESA), on the AWS Open Data registry, catalogued by Element 84's Earth Search, and TGE Labs' index of them on Source Cooperative: the Sentinel-2 finder.
• Nominatim and the OpenStreetMap Foundation, which find a place by its name in the Sentinel-2 finder.
• Terrain Tiles, from Mapzen, on AWS Open Data: the Relief background map.
• Natural Earth: the map in the drop zone.
• OpenFreeMap, by Zsolt Ero: the background map under every Geobox map, free, with no key and no limits.
• OpenStreetMap contributors, whose data that background map is drawn from, and OpenMapTiles for its schema.
• The Cloud-Native Geospatial Forum, where much of this comes together.