MCP4GVA

You run an MCP (Model Context Protocol) server that exposes the GVA GIS Soil Activities API through a simple, scriptable interface. It lets you query, count, and export features from the ArcGIS REST service you see at a fixed endpoint, so you can power MCP-enabled clients and automation with up-to-date GIS data.

How to use

Configure your MCP client to connect to the GVA GIS server endpoint and choose the operation you want to perform. You have multiple ways to run the server locally or remotely, and you can access four core tools to explore the layer, query features, count results, and export data to GeoJSON.

Recommended approach: run the MCP server locally if you are developing or testing, then switch to the hosted/MCP URL for production usage. The server exposes a remote HTTP endpoint you can call, and it also supports running as a local stdio server for tight integration with your tooling.

Key operations you can perform include: explore layer metadata, query features with SQL-like filters, count matching features, and export results to GeoJSON. You can combine these capabilities with paging and field selection to tailor responses to your needs.

How to install

# Prerequisite: install the uv runtime
# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# Windows
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

# Or install via pip
pip install uv

Option 1. Use uvx (recommended). Claude Desktop will download and run the MCP server automatically after you configure it.

Configure Claude Desktop for the built-in MCP server (example JSON): specify the uvx launcher and point to the MCP server name mcp4gva.

{
  "mcpServers": {
    "mcp4gva": {
      "command": "uvx",
      "args": ["mcp4gva"]
    }
  }
}

Option 2. Develop locally with uvx from a local path. Point Claude Desktop to your local repository path.

{
  "mcpServers": {
    "mcp4gva": {
      "command": "uvx",
      "args": [
        "--from",
        "/ruta/completa/a/mcp4gva",
        "mcp4gva"
      ]
    }
  }
}

Option 3. Traditional Python development with pip. Install in editable mode and configure Claude Desktop to run the module directly.

# Install in development mode
cd mcp4gva
pip install -e .

# Claude Desktop config
{
  "mcpServers": {
    "mcp4gva": {
      "command": "python",
      "args": ["-m", "mcp4gva.server"]
    }
  }
}

Additional sections

Project structure and standalone usage are provided for quick local testing. You can run a standalone Python client to verify access to the layer information, perform a query, count features, or export to GeoJSON.

from gva_gis_client import GVAGISClient

client = GVAGISClient()

# Get layer information
info = client.get_layer_info()

# Query features
result = client.query(where="1=1", result_record_count=10)

# Count features
count = client.count_features()

When working with Claude Desktop or another MCP client, you can also use the four exposed tools directly after configuration: gva_layer_info, gva_query, gva_count, and gva_export_geojson. They let you inspect the layer, fetch a subset of records, count matches, and export results as GeoJSON.

System and debugging notes: the server communicates via standard input/output streams. Logs appear on the console, making it straightforward to verify that the server is running and responding to requests.

Available tools

gva_layer_info

Fetches metadata about the layer, including fields, geometry type, spatial reference, and extent.

gva_query

Query features using SQL-like where clauses, select fields, and choose whether to include geometry. Supports pagination and field filtering.

gva_count

Count features that satisfy a where clause to quickly assess dataset size.

gva_export_geojson

Export matching features to GeoJSON with configurable fields and record limits.