Gdal Mcp

Name: Gdal Mcp
Author: Wayfinder-Foundry

@Wayfinder-Foundry

リポジトリ

Gdal Mcp について

About

#geospatial #gdal #raster #vector #eo #remote-sensing #geomatics #osgeo #gis

基本情報

カテゴリ

その他

トランスポート

stdio

公開者

Wayfinder-Foundry

投稿者

Jordan Gunn

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

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

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is Gdal Mcp?

Gdal Mcp is a Model Context Protocol server that gives AI agents geospatial analysis capabilities backed by GDAL, Rasterio, and PyProj. It requires AI agents to justify methodological choices—such as CRS selection or resampling—through a reflection middleware system, creating an auditable trail of epistemic reasoning. It is designed for developers and geospatial professionals who want reproducible, methodology-aware geospatial workflows with AI.

How to use Gdal Mcp?

Install via uvx --from gdal-mcp gdal --transport stdio or run with Docker. Configure the server in your MCP client (e.g., Claude Desktop) by adding a gdal-mcp entry to mcpServers with the uvx command and setting the GDAL_MCP_WORKSPACES environment variable. The server exposes 12 tools: 4 raster tools (info, convert, reproject, stats) and 6 vector tools (info, reproject, convert, clip, buffer, simplify), plus reflection system tools.

Key features of Gdal Mcp

Reflection middleware forces methodological justification before execution.
Structured justifications capture intent, alternatives, tradeoffs, and confidence.
Persistent cache achieves 75% hit rates across raster–vector workflows.
12 production-ready tools for raster and vector operations.
Workspace isolation and path validation for security.
Python-native stack (Rasterio, PyProj, pyogrio, Shapely, NumPy).

Use cases of Gdal Mcp

Reprojecting a DEM to UTM for accurate slope analysis with documented CRS reasoning.
Overlaying vector and raster datasets after justifying a common projection.
Converting file formats (GeoTIFF, COG, Shapefile, GeoJSON, GPKG) while recording methodological tradeoffs.
Clipping or buffering vector layers for spatial subsetting with an audit trail.
Educational scenarios where AI learns geospatial best practices through required justification.

FAQ from Gdal Mcp

What makes Gdal Mcp different from other AI geospatial tools?

It requires AI agents to justify methodological decisions (e.g., “why reproject to this CRS?”) before executing an operation, rather than silently acting. This prevents silent methodological errors like using nearest-neighbor resampling on continuous elevation data.

What dependencies and runtime are required?

Python 3.11+ with Rasterio, PyProj, pyogrio, Shapely, and NumPy. It runs via uvx from PyPI or Docker. FastMCP 2.0 provides the MCP server framework.

Where does my geospatial data live?

Data must be placed in workspaces specified by the GDAL_MCP_WORKSPACES environment variable. The server enforces workspace isolation and path validation to prevent access outside those directories.

What transport and authentication does it use?

The server communicates over stdio (standard input/output). No authentication mechanism is described; workspace isolation acts as the security boundary.

Are there any known limits?

The reflection system only triggers on a subset of tools (raster_reproject, vector_reproject). The cache is in-memory and is lost on server restart. Vector tools are new as of v1.1.1 and may have fewer features than raster tools.