STAC MCP Server provides access to STAC APIs for geospatial data discovery and access. It exposes a set of MCP tools that enable AI assistants and applications to interact with STAC catalogs: search and browse collections, retrieve metadata, and perform spatial and temporal queries. The server implements core tools including get_root, get_conformance, search_collections, get_collection, search_items, get_item, and estimate_data_size. Each tool supports an optional output_format parameter, defaulting to text, with JSON mode returning a compact envelope: { "mode": "json", "data": { ... } } to enable structured consumption (as per ADR 0006 and ASR 1003). The system also supports capability discovery and aggregations (ADR 0004) with graceful fallbacks when endpoints are missing, and describes how conformance and queryables behave in those scenarios. Data Size Estimation uses odc.stac to lazily load items into xarray datasets, clips AOI to the smallest common area when both bbox and AOI are provided, and returns detailed metadata about variables, coordinates, and assets. The MCP server can be run via Docker/Container and adheres to the MCP transport via stdio.