QGIS MCP

QGIS MCP bridges Claude Desktop and QGIS, the open-source GIS workhorse. You can ask Claude to create a project, load shapefiles or rasters, run processing algorithms, and execute PyQGIS code without touching the QGIS interface. It's a prompt-driven workflow for spatial data that feels surprisingly natural once you get past the setup. I'd reach for this when prototyping spatial analyses or teaching someone QGIS without forcing them through the GUI labyrinth. Ask Claude to load a boundary file, clip it to a region, calculate area, and export a styled map. The server handles the PyQGIS plumbing. It's faster than clicking through menus and documenting steps in a script at the same time. The install is fiddly. You need QGIS installed with a working Python environment, and the uv command assumes you've cloned the repo and are running from the right directory. The config snippet is listed as not applicable, which means you're piecing together the Claude Desktop setup from the repo README. Once it's wired up, though, it works. Claude can query available layers, run QGIS processing tools by name, and execute arbitrary PyQGIS. That last bit is powerful but also means you're trusting Claude to write correct spatial code. Quirks: QGIS must be running for the server to work, which isn't obvious from the name. It's not a standalone spatial engine, it's a remote control for an existing QGIS session. If you close QGIS, the MCP goes silent. Also, error messages from failed PyQGIS commands can be cryptic. Claude will try to fix them, but you'll want basic QGIS Python knowledge to debug. This isn't for casual users. If you don't already use QGIS or don't have spatial data workflows, the setup cost isn't worth it. But if you're a GIS analyst who wants to script workflows conversationally or a developer building spatial tools, it's a clever shortcut. It won't replace QGIS for production work, but it makes exploration and iteration much faster.