🌍 planetary-computer-mcp

Access petabytes of Earth observation data through natural conversation.

A Model Context Protocol (MCP) server that connects AI assistants to the Microsoft Planetary Computer — unlocking satellite imagery, climate data, land cover maps, and more. Query Sentinel-2, Landsat, NAIP aerial imagery, global DEMs, and 120+ geospatial datasets using simple spatial and temporal filters.

Sample Outputs

Add to VSCode, Cursor, Antigravity

Add the following to your mcp.json file (Cmd+Shift+P -> MCP: Add Server)

{
  "servers": {
    ...
    "planetary-computer": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "planetary-computer-mcp@latest"]
    },
  ...
  }
}

Install

npm install -g planetary-computer-mcp

# or locally
npx planetary-computer-mcp
bunx planetary-computer-mcp

Usage

Tools

• search_stac: Query STAC catalog by collection, bbox, datetime, limit
• get_collections: List all collections or get detailed info for a specific collection (assets, bands, resolutions)
• describe_collection: Get structured metadata with RGB/DEM/SAR strategy and recommended tools
• download_asset: Download GeoTIFF/assets with auto URL signing
• download_visual: Download RGB images with smart rendering (JPG for optical/DEM, PNG for classified)
• download_multispectral: Download specific bands into multi-band GeoTIFF
• download_geometries: Download vector data (e.g., MS Buildings) with spatial filtering
• download_zarr: Download spatial/temporal slices from Zarr collections (Daymet, ERA5, TerraClimate)
• render_zarr_preview: Create heatmap PNG previews from downloaded Zarr data

Supported Collections:

• Optical: sentinel-2-l2a, naip, landsat-c2-l2, hls2-l30/s30
• DEM: cop-dem-glo-30, alos-dem
• Land Cover: esa-worldcover, io-lulc-annual-v02, mtbs
• SAR: sentinel-1-rtc
• Vector: ms-buildings
• Climate/Weather: daymet-daily-na, daymet-daily-hi, daymet-daily-pr, era5-pds, terraclimate

Examples

• Search Sentinel-2: search_stac(collection="sentinel-2-l2a", bbox=[-122.5,47,-122,47.5], datetime="2024-06-01/2024-06-30")
• Visual Download: download_visual(collection="sentinel-2-l2a", bbox=[-122.4,47.6,-122.3,47.7], datetime="2024-06-01/2024-06-30")
• Multispectral: download_multispectral(collection="sentinel-2-l2a", assets=["B04","B08"], bbox=..., datetime=...)
• Buildings: download_geometries(collection="ms-buildings", bbox=[-122.35,47.6,-122.32,47.62])
• Climate Data: download_zarr(collection="daymet-daily-na", assets=["tmax","tmin"], bbox=[-122.5,47,-122,47.5], datetime="2024-01-01/2024-01-31")
• Zarr Preview: render_zarr_preview(zarr_path="path/to/daymet_daily_na/tmax")

MCP Integration

Add to client config:

{
  "mcpServers": {
    "planetary-computer": {
      "command": "npx",
      "args": ["-y", "planetary-computer-mcp"]
    }
  }
}

Architecture

• Language: TypeScript on Node.js ≥18
• Framework: MCP SDK with stdio transport
• API: Planetary Computer STAC catalog
• DuckDB: Spatial queries on parquet
• Features: Auto URL signing, streaming downloads, collection caching

Development

bun install               # Install deps
bun run build             # Compile to dist/
bun run format            # Format code
bun run check             # Typecheck, lint, prettier, tests
bun run test:integration  # Run all integration tests
bun run mcp               # Start MCP server