gdal-mcp

概述

GDAL MCP 是一个全面的模型上下文协议服务器，使 AI 代理能够使用 Rasterio、GeoPandas 和 PyProj 等 Python 库执行复杂的地理空间分析。其独特的反思中间件系统要求 AI 在执行前论证方法决策，创建关于'为什么'的对话，而不是仅仅执行'做什么'。这种方法防止了地理空间操作中的静默失败，记录了可重复性的方法，并通过跨操作重用论证实现了 75% 的缓存命中率。服务器提供 13 个生产就绪的工具，用于栅格和矢量数据操作，并有全面的文档。

试试问 AI

装完之后，这里有 5 个你可以让 AI 做的事：

• raster_info

  Inspect metadata of raster files (CRS, resolution, bands, nodata)

  试试:What's the CRS of this raster?试试:Show me the metadata for this DEM file
• raster_convert

  Convert raster formats with compression and overviews

  试试:Convert this GeoTIFF to COG format试试:Convert this raster to PNG with compression
• raster_reproject

  Transform raster to different CRS with reflection

  试试:Reproject this raster to Web Mercator试试:Change the projection of this elevation data
• raster_stats

  Calculate statistics and generate histograms for raster data

  试试:What are the statistics for this raster?试试:Generate a histogram for this image
• raster_query

  Perform spatial window queries on raster data

  试试:Extract values from this raster at these coordinates试试:Get the pixel values within this bounding box
• vector_info

  Inspect metadata of vector files (CRS, geometry, attributes)

  试试:What's the CRS of this shapefile?试试:Show me the schema of this vector layer
• vector_reproject

  Transform vector to different CRS with reflection

  试试:Reproject this vector to UTM zone 10N试试:Change the projection of these boundaries
• vector_convert

  Convert vector between different formats (SHP, GPKG, GeoJSON)

  试试:Convert this shapefile to GeoJSON试试:Change this GeoPackage to SHP format
• vector_clip

  Subset vector data spatially

  试试:Clip these boundaries to this extent试试:Extract features within this polygon
• vector_buffer

  Create buffer zones around vector geometries

  试试:Create a 1km buffer around these points试试:Buffer these lines by 500 meters
• vector_simplify

  Simplify geometries while preserving topology

  试试:Simplify these boundaries to reduce complexity试试:Reduce the detail in this coastline
• vector_query

  Perform spatial/attribute queries on vector data

  试试:Find all schools in this area试试:Query for features where name equals 'Park'

可对比工具

安装

# 直接从 PyPI 运行
uvx --from gdal-mcp gdal --transport stdio

{
  "mcpServers": {
    "gdal-mcp": {
      "command": "uvx",
      "args": ["--from", "gdal-mcp", "gdal", "--transport", "stdio"],
      "env": {
        "GDAL_MCP_WORKSPACES": "/path/to/your/geospatial/data"
      }
    }
  }
}

FAQ

GDAL MCP 与其他地理空间工具有什么不同？

GDAL MCP 独特的反思中间件要求 AI 在执行前论证方法选择，防止静默失败，并创建可重复的地理空间科学的文档化方法。

反思缓存系统如何工作？

系统缓存关于方法决策的认识推理，在稍后重复相似操作时实现 75% 以上的命中率。这种知识在不同数据类型（栅格/矢量）间持续存在以提高效率。

gdal-mcp 对比