A better spatial data infrastructure

Publish geospatial data
without the portal.

Portolan is a toolkit and set of conventions for publishing geospatial data as cloud-native files on object storage. STAC + GeoParquet + COG, done right, for governments and open data publishers.

// WHY PORTOLAN

Open, cheap, sovereign, AI-native.

Open

100% open source, open formats. Apache-2.0 license, open governance, open specs. If Portolan disappeared tomorrow, your data would still work everywhere.

GeoParquet · COG · STAC

AI-ready

STAC metadata is structured for retrieval. Catalogs are queryable by LLM agents and ready for semantic search — not just humans clicking through a portal.

STAC + RAG-friendly

Cheap

Pay only for storage and egress. No servers to run, no database to license, no portal seats to buy. A small government catalog: dollars a month.

storage + egress, that's it

Sovereign

Your data lives in your bucket, on your cloud, in your jurisdiction. Host on AWS, GCS, Azure, MinIO, Cloudflare R2 — anywhere S3-compatible.

bring your own bucket

Scales to petabytes

Cloud object storage handles the scale. Cloud-native formats serve range requests — a 100 GB COG opens in ~80ms, faster than most tile servers.

p95 < 120ms · global

Breaks the GIS silo

Query with DuckDB, Snowflake, BigQuery, Databricks, Pandas — not just GIS tools. Every layer is a file at a stable URL, same bytes for everyone.

one URL · every tool

// HOW IT WORKS

Files on storage, not servers.

Portolan replaces traditional SDI servers with cloud-native files on object storage. No databases, no tile servers, no portal software — just files that browsers and AI agents read directly.

Convert

Transform shapefiles, GeoTIFFs, and other formats into cloud-native equivalents — GeoParquet for vectors, COG for rasters, PMTiles for map tiles.

Catalog

Generate a STAC catalog that describes every asset with structured metadata, rich descriptions, and links that AI agents and tools can follow.

Publish

Push files to any S3-compatible object storage — AWS, GCS, Azure, R2, MinIO, or Source Cooperative for free open data hosting.

Browse

Data is instantly accessible through any STAC-compatible viewer, queryable via DuckDB, and ready for AI agents — no servers to run.

// THE TOOLKIT

The toolkit.

portolan-cli

The CLI

v0.7.0

Convert, validate, and sync. One command from a folder of shapefiles to a live STAC catalog on S3.

my-catalog · zsh

# Convert a folder of shapefiles + tiffs to a portable catalog

$ portolan ingest ./gov-data --to s3://my-catalog

✓ scanned 142 files (3.2 GB)

→ land_parcels.shp → GeoParquet (412 MB)

→ ortho_2024.tif → COG (2.1 GB)

→ roads_centerlines.shp → GeoParquet (84 MB)

→ ... 11 more

✓ STAC catalog generated (catalog.json + 14 collections)

✓ synced to s3://my-catalog (1.4 GB compressed)

▸ catalog.json: https://my-catalog.s3.amazonaws.com/catalog.json

▸ STAC browser: https://radiantearth.github.io/stac-browser/...

done · 0:48 elapsed

Works with AWS · GCS · Azure · R2 · MinIO · Source.coop

portolan-viewer

In design

Browser-native viewer

A lightweight viewer that reads STAC + cloud-native files directly.

portolan-skills

Stable

AI agent skills

Ready-made skills that let Claude, Gemini, and Codex agents publish catalogs and query cloud-native geospatial data.

// CATALOG REGISTRY

Browse 2 catalogs

Discover and connect to Portolan-compatible STAC catalogs from around the world.

Basic

A Portolan-managed STAC catalog

178 collections·STAC 1.1.0·34S, 60W

View catalog

Portolan NL — Cloud-Native Dutch Geodata