OpenLandMap MCP Server

Instructions

Implementation Reference

• src/openlandmap_mcp/tools/analysis.py:411-480 (handler)

  The function `build_r_snippet` implements the tool logic to generate R code snippets based on collection, item, asset keys, and an operation type. It is annotated with `@mcp.tool()` which registers it as an MCP tool.

  async def build_r_snippet(
      collection_id: str,
      item_id: str,
      asset_key: str,
      operation: str = "open",
  ) -> str:
      """Generate a ready-to-use R code snippet for accessing a raster asset.
  
      Uses terra, sf, and rstac packages.
  
      Args:
          collection_id: Collection identifier.
          item_id: Item identifier.
          asset_key: Asset key within the item.
          operation: Code operation:
              'open' — open with terra::rast()
              'info' — print metadata
              'plot' — plot the raster
              'clip_bbox' — crop to bounding box
              'stats' — compute statistics
              'export_csv' — export to CSV
  
      Returns:
          R code snippet as a string.
  
      Example:
          build_r_snippet("organic.carbon_usda.6a1c",
                          "organic.carbon_usda.6a1c_20180101_20181231",
                          "organic.carbon_usda.6a1c_m_1km_b30cm_s",
                          "plot")
      """
      data = await client.get_item_raw(collection_id, item_id)
      assets = data.get("assets", {})
  
      if asset_key not in assets:
          return f"# Error: asset '{asset_key}' not found. Available: {list(assets.keys())}"
  
      url = assets[asset_key].get("href", "")
  
      snippets = {
          "open": (
              f'library(terra)\n\n'
              f'url <- "{url}"\n'
              f'r <- rast(url)\n'
              f'print(r)\n'
              f'cat("CRS:", crs(r, describe=TRUE)$code, "\\n")\n'
              f'cat("Resolution:", res(r), "\\n")\n'
              f'cat("Extent:", as.vector(ext(r)), "\\n")'
          ),
          "info": (
              f'library(terra)\n\n'
              f'url <- "{url}"\n'
              f'r <- rast(url)\n'
              f'cat("Dimensions:", nrow(r), "x", ncol(r), "\\n")\n'
              f'cat("Layers:", nlyr(r), "\\n")\n'
              f'cat("CRS:", crs(r, describe=TRUE)$code, "\\n")\n'
              f'cat("Resolution:", res(r), "\\n")\n'
              f'cat("Extent:", as.vector(ext(r)), "\\n")\n'
              f'cat("Min:", minmax(r)[1], "\\n")\n'
              f'cat("Max:", minmax(r)[2], "\\n")\n'
              f'cat("Data type:", datatype(r), "\\n")'
          ),
          "plot": (
              f'library(terra)\n\n'
              f'url <- "{url}"\n'
              f'r <- rast(url)\n'
              f'plot(r, main="{collection_id} — {item_id}",\n'
              f'     col=hcl.colors(50, "viridis"))'
          ),
          "clip_bbox": (

• src/openlandmap_mcp/tools/analysis.py:410-410 (registration)

  The `@mcp.tool()` decorator on line 410 registers the `build_r_snippet` function as an MCP tool. The module `openlandmap_mcp.tools.analysis` is imported in `src/openlandmap_mcp/server.py` to ensure registration.

  @mcp.tool()