Microsoft Planetary Computer

The Microsoft Planetary Computer is a catalogue of environmental data that Microsoft hosts on its Azure cloud and makes available for free, with APIs for finding the data across space and time and a gallery of applications built on it by partners.² Its search API follows the SpatioTemporal Asset Catalog (STAC) standard, which describes where and when each satellite scene or data file was captured.² The service is still labelled a preview, and its terms say previews come without uptime guarantees, may change or be discontinued without notice, and are not meant for anything mission critical.^1,8

On 26 October 2026 the API listed 139 collections.⁴ They include the Landsat archive from 1972 to the present, Sentinel-2, Harmonized Landsat and Sentinel-2, MODIS, the ERA5 reanalysis, the Copernicus 30 m and 90 m elevation models, ESA WorldCover land cover, US NAIP aerial photography, Microsoft's building footprints and GBIF species records.^3,4 Each collection carries its own licence: ESA WorldCover is CC BY 4.0 and the building footprints are ODbL, for example.⁴

The STAC API can be used anonymously.^2,5 The files themselves sit in Azure Blob Storage, and reading them needs a short-lived token (a SAS token, a signed key appended to the file's address) that anyone can request without an account; the planetary-computer Python package fetches and refreshes these tokens automatically.⁵ Requests from registered users, or from inside Azure's West Europe region where most of the data is stored, get more generous rate limits and longer-lived tokens.^5,6 QGIS 4.0 and later can browse the catalogue directly through its built-in STAC browser.⁷ The APIs, the Python package and the catalogue website are open source on GitHub under the MIT licence.^2,12,14,15

Microsoft launched a hosted Jupyter notebook service, the Planetary Computer Hub, in April 2021 and retired it on 6 June 2024, citing company-wide security requirements the Hub could not meet; the data and APIs were unaffected.⁹ In its place Microsoft suggested Microsoft Fabric, Azure Machine Learning, GitHub Codespaces, or deploying one's own hub from its published guide.⁹ For production work Microsoft recommends the user's own computing, placed in West Europe next to the data.⁶ Microsoft Planetary Computer Pro is a separate Azure service for an organisation's own private geospatial data, deployed as a GeoCatalog resource in an Azure subscription; it reached general availability in June 2026.^10,11

For someone modelling a place, the Planetary Computer supplies satellite imagery, land cover and elevation for a watershed or county without downloading whole archives.^2,3 A script searches the STAC API for the area and dates, then reads only the files it needs.⁵ Unlike Google Earth Engine, it does not run the analysis itself, so the computing is the modeller's own.^6,9