TableGIS MCP Server

Geospatial data processing tools for AI assistants via the Model Context Protocol (MCP).

Wraps tablegis functions as MCP tools, enabling AI clients (Claude Desktop, Cursor, etc.) to perform spatial analysis through natural language.

Tools

Install

pip install tablegis-mcp

Or use with uvx (no install needed):

uvx tablegis-mcp

Configure

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "tablegis": {
      "command": "uvx",
      "args": ["tablegis-mcp"]
    }
  }
}

Or if installed via pip:

{
  "mcpServers": {
    "tablegis": {
      "command": "python",
      "args": ["-m", "tablegis_mcp.server"]
    }
  }
}

Claude Code

Add to ~/.claude.json (global) or project-level .claude/settings.json:

{
  "mcpServers": {
    "tablegis": {
      "command": "uvx",
      "args": ["tablegis-mcp"]
    }
  }
}

Or with pip:

{
  "mcpServers": {
    "tablegis": {
      "command": "python",
      "args": ["-m", "tablegis_mcp.server"]
    }
  }
}

Cursor / Other MCP Clients

Use the same command configuration. The server communicates via stdio transport.

Usage Examples

Once configured, you can ask your AI assistant:

• "Find the 3 nearest neighbors for each point in this dataset"
• "Draw 3km delivery zones around these store locations"
• "Calculate the area of each polygon in square metres"
• "Group points within 500m into clusters"
• "Match these coordinates to a shapefile of administrative boundaries"
• "Convert these WGS84 coordinates to GCJ-02 (Amap/高德)"

Data is passed as CSV or JSON strings; geometry results are returned as WKT.

Development

git clone https://github.com/Non-existent987/tablegis-mcp.git
cd tablegis-mcp
pip install -e ".[dev]"

License

MIT