Composable
geospatial infrastructure

From satellite to screen. No glue code.

Coming soon

claude mcp add folia -- ssh mcp@folia.sh

Example

One question, four layers

Specs compose together — not sequential steps, but ingredients that combine.

Question

"How much deforestation occurred in Colombian Amazon protected areas?"

→ 847,000 ha La Paya vs 1.2M in Tinigua

source

gfc: stac.umd/hansen-2023
protected: stac.wcmc/wdpa
bbox: [-75, -4, -70, 2]

pipeline

mask: treecover2000 > 30
extract: lossyear 2001-2023
buffer: 1km

analysis

zonal_stats:
  by: protected_area
  sum: loss_hectares

view

layout: split-map
layers: [loss, boundaries]
chart: comparison-bar

In action

Ask Claude, get maps

Once connected, folia tools are available in any conversation.

claude

you Identify reforestation sites in the Sierra Nevada based on slope, aspect, and burn severity

claude I'll query elevation and burn severity data for that region. ⟡ folia.query_sources bbox: [-121, 37, -119, 39], tags: ["dem", "mtbs"] Found USGS 3DEP and MTBS burn severity. Building the analysis pipeline. ⟡ folia.submit_pipeline pipeline: "analysis/reforestation-priority" ✓ Complete · 12,847 priority sites identified · View map →

you Generate viewshed analysis for these three proposed tower locations

claude I'll calculate visibility from each point using the local DEM. ⟡ folia.query_sources bbox: [-122.5, 37.7, -122.3, 37.9], type: "dem" Found 1m elevation data. Running viewshed from your three coordinates. ⟡ folia.submit_pipeline pipeline: "analysis/viewshed" ✓ Complete · 340 km² combined coverage · View results →

you Map evacuation routes avoiding slopes over 30% in Sonoma County

claude I'll combine road network data with terrain slope analysis. ⟡ folia.query_sources bbox: [-123.1, 38.2, -122.4, 38.8], tags: ["roads", "dem"] Found OpenStreetMap roads and USGS elevation. Filtering routes by slope threshold. ⟡ folia.submit_pipeline pipeline: "analysis/route-safety" ✓ Complete · 847 safe route segments · View network →

Capabilities

Built for the modern geo stack

Cloud-native formats, federated catalogs, run anywhere.

◇

Federated catalog

Index data where it lives. STAC pointers to Planetary Computer, AWS, your own buckets. Never duplicate.

⟳

Run anywhere

Same pipeline runs locally or as parallelized batch compute. Develop on your laptop, scale to the cloud.

Cost-aware

Estimate before you execute. Transparent pricing, compare providers, no surprise bills.

≡

Cache by content

Same inputs + steps = cached output. Content-addressed storage. Never recompute what already exists.

☁

Cloud-native output

PMTiles, COG, GeoParquet. Optimized for streaming, not downloading. HTTP range requests everywhere.

⬡

MCP-native

Expose everything via Model Context Protocol. Any AI client can query, estimate, and execute pipelines.

Architecture

The composable stack

Four layers, each outputting standard formats. Use the full stack or eject at any point.

Source

Federated data catalogs

STAC COG Parquet

Pipeline

Transform and process

slope reproject tile

Analysis

Compute on demand

zonal viewshed query

View

Render anywhere

MapView Chart Card

Composablegeospatial infrastructure