STAC MCP Server

Overview

The STAC MCP Server bridges AI assistants with the SpatioTemporal Asset Catalog (STAC) ecosystem, enabling seamless discovery and retrieval of geospatial datasets. By exposing STAC endpoints through the Model Context Protocol, developers can ask an AI to search for satellite imagery, weather products, or any other asset cataloged in STAC without writing custom HTTP code. The server translates MCP tool calls into standard STAC API requests and returns results in either human‑readable text or a structured JSON envelope, ensuring compatibility with both legacy clients and modern workflows that prefer machine‑processable data.

What sets this server apart is its dual output mode. Every tool accepts an optional parameter; the default produces concise, chat‑friendly summaries, while returns a compact JSON envelope that preserves the full payload. This design allows AI assistants to switch between conversational explanations and structured responses on demand, making the same tool usable in a chat interface or as part of an automated data pipeline.

Key capabilities include:

• Catalog navigation (, , ) for browsing available collections and their metadata.
• Item discovery (, ) with support for spatial, temporal, and attribute filters.
• Queryable field inspection () to understand which attributes can be used in filters.
• Aggregations () for quick statistics such as counts or min/max values without downloading full data.
• Data size estimation () that leverages lazy loading to predict download sizes, helping users plan bandwidth and storage.

The server also implements robust capability discovery. If a catalog lacks certain endpoints—such as or —the tools return structured JSON indicating the feature is unsupported rather than raising hard errors. This graceful fallback lets AI clients adapt their behavior dynamically, for example by falling back to the root document’s array when is missing.

In real‑world scenarios, the STAC MCP Server empowers developers to build AI‑driven geospatial applications: an assistant can answer “Show me all MODIS imagery over the Amazon between 2022‑01‑01 and 2022‑12‑31,” fetch the relevant items, estimate their download size, and even provide aggregated statistics—all through simple MCP tool calls. By encapsulating STAC interactions behind a standard protocol, the server removes the need for custom SDKs or REST wrappers, accelerating prototype development and fostering interoperability across AI platforms.