Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Planetary Computer MCP Server show me sentinel-2 imagery of San Francisco"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A Python implementation of the Planetary Computer MCP server, providing unified access to satellite and geospatial data through natural language queries.
Related MCP server: Google Earth Engine MCP Server
Unified Interface: Single download_data tool that automatically detects datasets from natural language queries
Natural Language Geocoding: Automatically converts place names (e.g., "San Francisco", "the Alps", "Amazon rainforest") to geospatial bounding box coordinates using the Nominatim geocoding service—no need to manually specify coordinates
Multi-format Support: Raster (GeoTIFF), Vector (GeoParquet), and Zarr data
Automatic Visualization: Generate RGB/JPEG previews for LLM analysis
Fast Downloads: Uses odc-stac for efficient COG access
uv syncpython -m planetary_computer_mcp.serverfrom planetary_computer_mcp.tools.download_data import download_data
# Download Sentinel-2 data for San Francisco
result = download_data(
query="sentinel-2 imagery",
aoi="San Francisco",
time_range="2024-01-01/2024-01-31"
)
print(f"Raw data: {result['raw']}")
print(f"Visualization: {result['visualization']}")Unified tool for raster, DEM, land cover, and climate data.
Parameters:
query: Natural language query (e.g., "sentinel-2", "elevation data")
aoi: Bounding box [W,S,E,N] or place name
time_range: ISO8601 datetime range
max_cloud_cover: Maximum cloud cover (optical data)
Returns:
Raw GeoTIFF/Zarr/Parquet file
RGB/JPEG visualization
Metadata
Tool for vector/building data.
Parameters:
collection: Collection ID (e.g., "ms-buildings")
aoi: Bounding box or place name
limit: Maximum features
Returns:
GeoParquet file
Map visualization
Feature count
See collections.md for the complete list of supported datasets.
uv sync --devuv run pytestuv run pre-commit run --all-filessrc/
├── core/ # Core utilities
│ ├── stac_client.py # STAC search wrapper
│ ├── geocoding.py # Place name → bbox
│ ├── collections.py # Dataset metadata
│ ├── raster_utils.py # odc-stac helpers
│ ├── vector_utils.py # DuckDB helpers
│ ├── visualization.py # Matplotlib viz
│ └── zarr_utils.py # Xarray Zarr helpers
├── tools/ # MCP tools
│ ├── download_data.py
│ └── download_geometries.py
└── server.py # MCP server entry pointApache 2.0 License
Download satellite/raster data from Microsoft Planetary Computer.
Automatically detects collection from natural language queries, handles geocoding for place names, downloads and crops data, generates RGB visualizations.
query : str Natural language query describing the data you want. Examples: "sentinel-2 imagery", "landsat", "naip aerial photos", "elevation data", "land cover" aoi : str or list[float] Required. Area of interest as either: - Place name string: "Seattle, WA", "Paris, France", "Central Park, NY" - Bounding box list: [west, south, east, north] in degrees Example: [-122.4, 47.5, -122.3, 47.6] time_range : str or None, optional ISO8601 datetime range. Defaults to last 7 days if not provided. Examples: "2024-01-01/2024-01-31", "2024-06-01/2024-06-30" output_dir : str, optional Directory to save outputs. Defaults to current directory. max_cloud_cover : int, optional Maximum cloud cover percentage for optical data (0-100). Default: 20
list[TextContent] File paths and metadata
Download recent Sentinel-2 imagery of Paris: query="sentinel-2 imagery", aoi="Paris, France"
Download Landsat for a specific bbox and time: query="landsat", aoi=[-122.4, 47.5, -122.3, 47.6], time_range="2024-06-01/2024-06-30"
Download NAIP aerial imagery: query="naip aerial photos", aoi="Central Park, NY"
Inputs: query (required), aoi (required), time_range, output_dir, +1 more. Returns: result
Download vector geometries from Planetary Computer.
collection : str Collection ID (e.g., "ms-buildings") aoi : list[float] or str Bounding box [W,S,E,N] or place name string output_dir : str, optional Directory to save outputs
list[TextContent] File paths and metadata
Inputs: collection (required), aoi (required), output_dir. Returns: result
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 2 tools
Both tools have clearly distinct purposes: one for raster satellite data with natural language querying, and one for vector geometries with a collection ID. No overlap in functionality.
Both tools follow a consistent download_<type>_tool pattern using snake_case, making the naming predictable and clear.
With only 2 tools, the server feels very thin for the broad scope of Microsoft Planetary Computer. Even a minimal interface would benefit from additional tools like listing collections or searching items.
The server covers raster and vector data downloads but lacks critical supporting operations such as querying available collections, filtering items, or exploring data metadata. Users must know collection IDs or rely on natural language detection without discovery tools.
Geospatial AI MCP server — satellite imagery, embeddings, weather, GNS governance
One MCP server for all web data. Scrape, search, crawl, and AI research in one install.
Hosted MCP server for live public-data APIs and Skills for AI agents.
