MCP servers
Geospatial MCP servers
Hand-reviewed Model Context Protocol servers that let AI agents query spatial databases, run GDAL and QGIS, geocode, and build maps.
Latest servers
View allarcgis-portal-mcp
ArcGIS
MCP server with 81 tools for ArcGIS Enterprise Portal and ArcGIS Online: search, query, edit, publish, geoprocess, and admin via the REST API.

qgis_mcp (jjsantos01)
Geoprocessing
Early QGIS Desktop to Claude MCP integration via PyQGIS.

OSM-GeoJSON-MCP-Server
OpenStreetMap
Extract OSM buildings, roads, amenities, water, green and rail as GeoJSON via Overpass.

QgisRemoteMCP
GDAL
Full QGIS Desktop in Docker via noVNC as an HTTP MCP (1000+ Processing algorithms).

arcgis-mcp-bridge
ArcGIS
~100 ArcGIS Pro / ArcPy tools over stdio with sandboxing.

postgres-postgis-docsbox-mcp
PostGIS
Read-only PostGIS MCP with a SQL validator, introspection and a PostGIS docs corpus.
Felt MCP Server
Cartography
Hosted Felt GIS MCP (~30 tools): maps, warehouse SQL, spatial analysis, styling and collaboration.
CARTO MCP Server
Cartography
Remote CARTO platform MCP: warehouse explore, Builder maps, Workflows, spatial SQL, geocode/route and admin.
Browse by workflow
Popular searches
About the GIS MCP server directory
This directory brings together AI tooling built for geospatial work. MCP servers expose GIS capabilities — querying a PostGIS database, running GDAL, driving QGIS, geocoding, or working with STAC imagery — to an AI agent over the Model Context Protocol. Agent skills package instructions loaded directly into an agent's context, and plugins extend an assistant with a specific function. Filter by category or search by name to find the right server for your workflow.
Frequently asked questions
Everything you need to know about connecting GIS capabilities to AI agents.
Have a GIS MCP server or skill to add?
Contribute it to the directory so other engineers can find and use it with their AI agents.
Contribute to the directoryarrow_forwardWhat is a GIS MCP server?expand_more
An MCP (Model Context Protocol) server is a program that exposes GIS capabilities — such as querying a PostGIS database, running GDAL operations, or driving QGIS — to an AI agent through a standard protocol. Agents like Claude, Cursor, or VS Code connect to the server and call its tools to perform geospatial work.
What is the difference between an MCP server, an agent skill, and a plugin?expand_more
An MCP server is a standalone process an AI agent connects to over a transport such as stdio or HTTP. An agent skill is a packaged instruction set and toolset loaded directly into an agent's context. An agent plugin — such as GeoSQL — is a packaged capability or connector that extends an AI assistant with a specific function, like running spatial SQL, invoked through the assistant itself.
How do I install one of these tools?expand_more
Each tool's detail page includes a copyable install block with the exact configuration for its type — the command and arguments for an MCP server, the package reference for a skill, or the setup steps for a plugin — along with the supported transport and compatible agent targets.
Are these GIS AI tools free?expand_more
The directory lists open-source and freely available GIS tooling. Each entry links to its homepage and source repository and records its license so you can confirm the terms before use.
What is an agent skill?expand_more
An agent skill is a packaged set of instructions and reference material loaded directly into an AI agent's context. It teaches the agent how to approach a task — for example, writing performant PostGIS queries or structuring a spatial schema — without running a separate process.
How is an agent skill different from an MCP server?expand_more
An MCP server is a running program the agent connects to over a transport such as stdio or HTTP; it exposes callable tools. An agent skill is static knowledge loaded into the agent's context — no process, no transport. A skill shapes how the agent reasons; an MCP server gives the agent new actions to call.
How do I install an agent skill?expand_more
Each skill's detail page shows a single copyable install line — typically a command that adds the skill from its source repository to your agent. There is no archive to download; the skill is pulled directly from its repository.