Geospatial and mapping is one of the richest MCP categories we’ve reviewed. AI agents that can geocode addresses, calculate routes, process satellite imagery, perform spatial analysis, and generate maps through natural language unlock workflows that previously required specialized GIS training. The category now spans six areas: commercial mapping platforms (Mapbox, Google Maps, Baidu Maps, TomTom, HERE Maps), earth observation & remote sensing (NASA Earthdata, Google Earth Engine, Axion Planetary), open-source mapping (OpenStreetMap, QGIS), GIS operations libraries (gis-mcp, GeoServer, LocuSync), government geospatial data (Japan MLIT), and data conversion tools.
The headline finding: QGIS MCP surged to 926 stars — the most popular geospatial MCP server by far, gaining 13 stars in 8 days despite no new code. Google Maps community server (cablate) hit 279 stars with v0.0.52 adding configurable HTTP bind interface for remote access and improved transit direction handling. TomTom rebranded to “TomTom Maps MCP Server” — the rename disambiguates from a separate Traffic Analytics MCP product, signaling TomTom is expanding its MCP portfolio. Mapbox continues with two official servers — the main server (335 stars) with hosted endpoint at mcp.mapbox.com, and the DevKit (49 stars) which upgraded to OpenTelemetry v2.x and consolidated docs to 4 core static resources. Axion Planetary holds at 218 stars with AWS-hosted SAR-to-optical satellite imagery and 935 commits. Japan’s MLIT servers both grew — dpf to 155 stars, geospatial to 170 stars. The main gap remains that Google Maps has no official MCP server — though cablate’s community option is increasingly comprehensive.
Category: Science & Research
Commercial Mapping Platforms
Mapbox
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| mapbox/mcp-server | 335 | TypeScript | 20 | stdio, hosted |
mapbox/mcp-server (335 stars, TypeScript, MIT, 121 commits) is the official MCP server from Mapbox. Twenty tools organized into three categories:
Offline geospatial utilities (8 tools) — distance, point_in_polygon, bearing, midpoint, centroid, area, bounding_box, buffer, simplify. Local calculations using Turf.js — no API calls needed. These work without a Mapbox token for purely geometric operations.
API-powered tools (10 tools) — search_and_geocode, reverse_geocoding, category_search, directions, matrix, isochrone, map_matching, optimization, static_image. Full access to Mapbox’s geocoding (global address resolution), routing (driving/walking/cycling with turn-by-turn), travel time matrices (up to 25×25 origins/destinations), isochrones (reachability polygons), map matching (snap GPS traces to roads), and route optimization (multi-stop TSP solving).
Resource tools (2 tools) — resource_reader for fetching Mapbox tilesets and datasets, plus a deprecated category_list tool.
Standout feature: hosted endpoint at mcp.mapbox.com/mcp — connect directly without installing or running anything locally. Just add the URL and your Mapbox access token. This is the simplest setup of any geospatial MCP server. Supports Claude Desktop, VS Code, Cursor, and the Goose framework.
Requires a Mapbox access token (free tier available with generous limits). The combination of zero-install hosted mode, offline geometric tools, and comprehensive API coverage makes this the best starting point for geospatial MCP integration.
mapbox/mcp-devkit-server (49 stars, TypeScript, 195 commits) is Mapbox’s second official MCP server, focused on developer workflows rather than location data queries. Tools include style creation and management (ListStylesTool, ValidateStyleTool), access token generation, GeoJSON formatting and visualization (generates geojson.io URLs for instant preview), coordinate reprojection, and bounding box calculation. A hosted version is available for zero-install setup. This complements the main MCP server — use the main server when agents need geocoding/routing/POI data, use the DevKit when agents are building or debugging Mapbox applications. Supports Claude Desktop, Claude Code, VS Code, and Cursor.
Google Maps
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| cablate/mcp-google-map | 279 | TypeScript | 18 | stdio, HTTP |
| david-pivonka/google-maps-mcp-server | 2 | TypeScript | 14 | stdio |
Google has no official Google Maps MCP server. Two community servers fill the gap:
cablate/mcp-google-map (279 stars, TypeScript, MIT, 118 commits) is the most popular and continues growing steadily. Now at v0.0.52 with 18 tools (14 atomic + 4 composite). Recent updates: v0.0.52 (April 21) added a configurable HTTP bind interface via CLI argument and environment variable, defaulting to all interfaces for easier remote access. v0.0.51 (April 19) improved error messaging for transit directions in unsupported regions and fixed timestamp validation with the Routes API. Tools include maps_air_quality, maps_static_map, maps_batch_geocode, maps_search_along_route, and maps_local_rank_tracker (geographic grid-based ranking analysis for local SEO — unique among mapping MCP servers). The 4 composite tools include maps_explore_area, maps_plan_route, maps_compare_places, and maps_local_rank_tracker. Runs in three modes: stdio, StreamableHTTP server, and standalone CLI execution.
david-pivonka/google-maps-mcp-server (2 stars, TypeScript, MIT, 15 commits) has broader API coverage with 14 tools including places_autocomplete, places_photos, roads_nearest, ip_geolocate, and geolocation_estimate — APIs not covered by cablate’s server. More niche tools but less adoption and no composite operations.
Both require a Google Cloud API key with the Places API (New) enabled. Google also offers the Google Maps Platform Code Assist toolkit — an MCP server focused on documentation and code samples rather than live API access, aimed at developers building Maps apps rather than agents using Maps data.
Baidu Maps
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| baidu-maps/mcp | 411 | Python/JS | 10 | stdio, SSE |
baidu-maps/mcp (422 stars, Python 55.9%/JavaScript 44.1%, MIT, 67 commits) is the official Baidu Maps MCP server — the first map service provider in China to support MCP. Ten tools: map_geocode, map_reverse_geocode, map_search_places, map_place_details, map_directions_matrix, map_directions, map_weather, map_ip_location, map_road_traffic, map_poi_extract.
The most notable inclusion is map_road_traffic for real-time traffic conditions and map_poi_extract for extracting points of interest from text — capabilities not found in most mapping MCP servers. Dual SDK support (Python and JavaScript/TypeScript). Requires a server-side API key from Baidu Maps Open Platform with MCP (SSE) service enabled.
Primary value: comprehensive Chinese mapping data. If your use case involves Chinese addresses, roads, or POIs, this is the only real option — Google Maps coverage in China is limited and Mapbox requires separate China-specific tokens.
TomTom
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| tomtom-international/tomtom-mcp | 46 | TypeScript | 18 | stdio, HTTP |
tomtom-international/tomtom-mcp (46 stars, TypeScript, Apache 2.0, 608 commits) is the official TomTom MCP server, recently rebranded to TomTom Maps MCP Server to disambiguate from a separate Traffic Analytics MCP product — signaling TomTom is expanding its MCP portfolio. Eighteen tools across two backends:
Standard tools (11) — tomtom-geocode, tomtom-reverse-geocode, tomtom-fuzzy-search, tomtom-poi-search, tomtom-nearby, tomtom-routing, tomtom-waypoint-routing, tomtom-reachable-range, tomtom-traffic, tomtom-static-map, tomtom-dynamic-map. Available on both TomTom Maps and Orbis Maps backends.
Orbis Maps exclusive (5) — tomtom-ev-routing, tomtom-search-along-route, tomtom-area-search, tomtom-ev-search, tomtom-data-viz. The EV routing tools are unique — they calculate routes accounting for battery range, charging station locations, and energy consumption, something no other mapping MCP server offers.
The 608-commit count is the highest of any mapping MCP server, indicating serious ongoing development. Now offers a remote HTTP endpoint for zero-install access alongside local stdio mode — join Mapbox and DigitalOcean in the small club of hosted geospatial MCP servers. Includes a debug UI for visually testing tools with interactive map widgets. Requires Node.js 22.x and a TomTom API key (free developer tier available).
HERE Maps
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| limingchina/heremaps-mcp-server | 8 | JavaScript | 6 | stdio |
limingchina/heremaps-mcp-server (6 stars, JavaScript, Apache 2.0, 14 commits, inactive since January 2025) is a community server for HERE Maps. Six tools: maps_geocode, maps_reverse_geocode, maps_search_places, maps_directions, maps_get_traffic_incidents, maps_display. Basic but functional coverage of HERE’s core APIs. Requires a HERE Maps API key. No official server from HERE exists.
Earth Observation & Remote Sensing
NASA Earthdata
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| nasa/earthdata-mcp | — | Python | 2+ | stdio |
| datalayer/earthdata-mcp-server | 23 | Python | 2+ | stdio |
| ProgramComputer/NASA-MCP-server | 83 | TypeScript | 20+ | stdio |
Three servers bring NASA data to AI agents:
nasa/earthdata-mcp (Python, updated March 2026) is NASA’s official MCP server for Earthdata. It provides semantic search powered by embeddings for discovering Earth science datasets through NASA’s Common Metadata Repository (CMR). Core tools include search_earth_datasets and search_earth_datagranules for finding datasets and granules within specific collections. The architecture uses AWS Lambda with Terraform infrastructure-as-code, suggesting this is designed for cloud deployment. Each tool is self-contained in its own folder with a manifest and implementation file. This is the first official NASA MCP server — a significant signal for the geospatial community.
datalayer/earthdata-mcp-server (25 stars, Python, 32 commits) is a community Earthdata server that adds Jupyter notebook integration — it composes all Jupyter MCP Server tools alongside Earth data discovery, providing a unified interface for finding NASA datasets and analyzing them in notebooks. Useful for researchers who want dataset discovery and interactive analysis in one workflow.
ProgramComputer/NASA-MCP-server (83 stars, TypeScript, MIT, 31 commits, npm package @programcomputer/nasa-mcp-server) takes a broader approach, covering 20+ NASA APIs through a single interface: APOD (Astronomy Picture of the Day), Mars Rover Photos, EPIC (Earth Polychromatic Imaging Camera), NEO (Near Earth Object Web Service), EONET (Earth Observatory Natural Event Tracker), DONKI (Space Weather Database), NASA Image and Video Library, and more. Output is normalized for ML model ingestion. While not focused on Earthdata specifically, it provides the widest coverage of NASA’s public APIs.
Google Earth Engine
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| Dhenenjay/axion-planetary-mcp | 217 | TypeScript | 30+ | stdio, SSE |
| cameronking4/google-earth-engine-mcp | — | TypeScript | — | stdio |
Google Earth Engine (GEE) provides petabytes of satellite imagery and geospatial datasets. Two community MCP servers unlock this for AI agents:
Dhenenjay/axion-planetary-mcp (218 stars, TypeScript, 935 commits) has nearly doubled its stars since our last review and released V2.0 — now powered by AWS. The migration from Google Cloud to AWS brings improved performance and global accessibility. A hosted server is now available at https://axion-mcp-sse.onrender.com for zero-install access. The exclusive axion_sar2optical feature converts SAR radar imagery to optical-like images, enabling cloud-free, day/night Earth observation. Over 30 tools and 5 pre-trained models, including the TerraMind encoder + DARN adaptive decoder architecture (86.66% mIoU accuracy). Capabilities include NDVI/NDWI vegetation analysis, crop classification, wildfire risk assessment, deforestation tracking, and interactive map generation. Supports Sentinel-2, Landsat, and MODIS satellite datasets. ~7,458 monthly npm downloads. The 935-commit count and rapid star growth make this the fastest-growing server in the geospatial category.
cameronking4/google-earth-engine-mcp (TypeScript) enables natural language queries against GEE — search datasets, calculate vegetation indices, filter collections by location and date, run tasks, export imagery to cloud storage, and visualize results in chat. A simpler alternative to Axion for straightforward GEE workflows. Requires a GEE service account JSON key.
Neither server is official from Google. Google Earth Engine itself has no MCP server, making this a significant community-filled gap — GEE is arguably the most important geospatial data platform for environmental science.
Open-Source Mapping
QGIS
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| jjsantos01/qgis_mcp | 926 | Python | 7 | stdio |
jjsantos01/qgis_mcp (926 stars, Python) connects QGIS Desktop to AI assistants through MCP. A major Tool Overhaul in v1.0.1 streamlined the server from 36 tools down to 7 consolidated “super tools” — this reduces context window usage and improves LLM accuracy in selecting the right tool. The remaining tools pack more functionality into each call.
The standout remains execute_processing — it exposes QGIS’s entire processing algorithm library (hundreds of geoprocessing tools) through a single MCP tool. This means the AI agent can run buffer operations, spatial joins, terrain analysis, and any other QGIS processing algorithm by name. Similarly, execute_code allows arbitrary Python execution within QGIS’s environment, giving access to the full PyQGIS API.
This is a desktop integration — it requires a running QGIS 3.x instance with the MCP plugin installed. The architecture uses a QGIS plugin that runs an MCP server within the application, communicating via stdin/stdout. Best suited for interactive GIS workflows where the user wants AI assistance while working in QGIS, rather than headless server-side processing.
The 926-star count makes it the most popular geospatial MCP server by a wide margin — GIS professionals clearly want AI integration in their existing tools.
OpenStreetMap
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| jagan-shanmugam/open-streetmap-mcp | 187 | Python | 12 | stdio |
| wiseman/osm-mcp | 80 | Python | 7 | stdio |
| NERVsystems/osmmcp | 20 | Go | 25 | stdio |
Three OpenStreetMap MCP servers with different approaches:
jagan-shanmugam/open-streetmap-mcp (187 stars, Python, MIT, 9 commits) is the most popular. Twelve tools focused on practical location queries: geocode_address, reverse_geocode, find_nearby_places, get_route_directions, search_category, suggest_meeting_point, explore_area, find_schools_nearby, analyze_commute, find_ev_charging_stations, analyze_neighborhood, find_parking_facilities. The specialized tools (schools, EV charging, parking, commute analysis) make this feel more like a location assistant than a raw geocoding API. Uses Nominatim and OSRM — no API key required, though rate limits apply.
wiseman/osm-mcp (80 stars, Python/HTML, 21 commits) takes a different approach — it combines map visualization with direct database queries. Seven tools: get_map_view, set_map_view, set_map_title, add_map_marker, add_map_line, add_map_polygon, query_osm_postgres. Includes a web-based Leaflet map viewer with Server-Sent Events for real-time updates. Requires a PostgreSQL/PostGIS database loaded with OpenStreetMap data — heavier setup but allows complex spatial SQL queries against the full OSM dataset.
NERVsystems/osmmcp (20 stars, Go, MIT, 173 commits) has the most tools (25) and most active development. Adds emission enrichment (enrich_emissions), polyline encoding/decoding, bbox calculations, tag filtering, and distance sorting on top of the standard geocoding/routing/POI capabilities. Written in Go for performance. The 173-commit count suggests ongoing maintenance.
All three are free — OpenStreetMap data requires no API key. The tradeoff is rate limits on Nominatim/OSRM public instances vs. running your own.
GIS Operations Libraries
gis-mcp
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| mahdin75/gis-mcp | 139 | Python | 100+ | stdio, HTTP/SSE |
mahdin75/gis-mcp (139 stars, Python, 249 commits, v0.14.0 beta) is the most comprehensive GIS MCP server. Over 100 tools bridging six Python GIS libraries to LLMs:
Shapely (29 tools) — buffer, intersection, union, difference, symmetric difference, convex hull, simplify, and other geometric operations. The core computational geometry toolkit.
PyProj (13 tools) — coordinate transformations between any CRS. Reproject geometries, transform coordinate arrays, compute geodesic distances and areas on the ellipsoid.
GeoPandas (13 tools) — read/write geospatial files (Shapefile, GeoJSON, GeoPackage), spatial joins, overlays, dissolves, and attribute operations on geodataframes.
Rasterio (20 tools) — raster data I/O, clipping, reprojection, NDVI computation, hillshade generation. Bridges the vector/raster divide.
PySAL (18 tools) — spatial statistics, spatial weights, spatial autocorrelation (Moran’s I, LISA), and regionalization algorithms.
Visualization (2 tools) — create_static_map and create_interactive_web_map for rendering results.
Data access (6 tools) — administrative boundaries, climate data, ecology data, movement data, land cover, and satellite imagery retrieval.
This is effectively a complete GIS workstation exposed via MCP. The breadth is unmatched — no other server covers geometry operations, coordinate transforms, raster processing, spatial statistics, and data visualization in one package. Now supports HTTP/SSE transport for network deployment and Docker containerization. Includes integration with AI frameworks (LangChain, OpenAI) and file storage endpoints. Requires Python 3.10+. Published on PyPI as gis-mcp.
GeoServer MCP
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| mahdin75/geoserver-mcp | 70 | Python | 40+ | stdio |
mahdin75/geoserver-mcp (70 stars, Python, 42 commits) connects LLMs to the GeoServer REST API. Over 40 tools across workspace management (create/list workspaces), datastore management (Shapefile, GeoPackage, PostGIS datastores), layer management (create, configure, list, remove layers), layer group management (7 tools), user and access control (7 tools), feature type and attribute management (6 tools), style management, and system operations.
By the same author as gis-mcp — together they cover both GIS analysis (gis-mcp) and geospatial data serving (geoserver-mcp). Requires a running GeoServer instance with REST API enabled. Python 3.10+.
LocuSync (GIS-MCP-Server)
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| matbel91765/gis-mcp-server | 1 | Python | 14 | stdio |
matbel91765/gis-mcp-server (1 star, Python, MIT, 32 commits) provides geocoding, routing, elevation, and spatial analysis using free public APIs (Nominatim, OSRM, Open-Elevation). Fourteen tools: geocode, reverse_geocode, batch_geocode, get_elevation, get_elevation_profile, distance, buffer, spatial_query, transform_crs, route, isochrone, read_file, write_file. File I/O supports Shapefile, GeoJSON, and GeoPackage formats.
A good lightweight alternative if you need basic geospatial operations without the heavy dependency chain of gis-mcp. No API keys required.
Government Geospatial Data
Japan MLIT
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| MLIT-DATA-PLATFORM/mlit-dpf-mcp | 155 | Python | 18 | stdio |
| chirikuuka/mlit-geospatial-mcp | 170 | Python | 1 | stdio |
Two servers for Japan’s Ministry of Land, Infrastructure, Transport and Tourism (MLIT) data:
MLIT-DATA-PLATFORM/mlit-dpf-mcp (155 stars, Python, MIT, 4 commits) is the official MLIT Data Platform MCP server. Eighteen tools for searching, browsing, and downloading Japan’s national geospatial datasets through GraphQL APIs: search, search_by_location_rectangle, search_by_location_point_distance, search_by_attribute, get_data, get_data_summary, get_data_catalog, get_file_download_urls, get_prefecture_data, get_municipality_data, get_mesh, and more. Covers infrastructure, transportation, land use, and environmental datasets.
chirikuuka/mlit-geospatial-mcp (170 stars, Python, MIT, 16 commits) focuses on the Real Estate Information Library API — a single get_multi_api tool that accesses 25 of 35 available data types covering property prices, disaster risk zones, schools, medical facilities, population projections, and urban planning designations. Requires an API key from the Real Estate Information Library.
These are niche but notable — Japan is ahead of most countries in providing official MCP access to government geospatial data.
Data Conversion
GIS Data Conversion MCP
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| ronantakizawa/gis-dataconversion-mcp | 15 | JavaScript | 9 | stdio |
ronantakizawa/gis-dataconversion-mcp (15 stars, JavaScript, MIT, 19 commits) converts between geospatial formats. Nine tools: wkt_to_geojson, geojson_to_wkt, csv_to_geojson, geojson_to_csv, geojson_to_topojson, topojson_to_geojson, kml_to_geojson, geojson_to_kml, coordinates_to_location. Over 1,000 npm downloads. Simple and focused — when you just need to convert between WKT, GeoJSON, CSV, TopoJSON, and KML formats.
Desktop GIS Integration
ArcGIS Pro
| Server | Stars | Language | Tools | Transport |
|---|---|---|---|---|
| nicogis/MCP-Server-ArcGIS-Pro-AddIn | 31 | C# | 4 | stdio (Named Pipes) |
nicogis/MCP-Server-ArcGIS-Pro-AddIn (31 stars, C#, 10 commits) integrates MCP with ArcGIS Pro Desktop. Four tools: pro.getActiveMapName, pro.listLayers, pro.countFeatures, pro.zoomToLayer. The architecture is clever — a .NET Add-In runs inside ArcGIS Pro and communicates with a separate .NET 8 MCP server via Named Pipes. Currently read-only with limited tools, but demonstrates the pattern for AI-assisted desktop GIS. Requires Visual Studio 2022, ArcGIS Pro SDK, and ArcGIS Pro installation.
Also notable
- Google Maps Platform Code Assist — an official Google MCP server focused on documentation and code samples for developers building Maps applications, not for live API access
- Cesium AI Integrations — a repository (March 2026) exploring MCP server integration with Cesium’s 3D geospatial platform; still early
- ThinkGeo MCP Server — indexes 12,700+ ThinkGeo documentation pages for developer assistance; no API key required
- Baidu Maps community (hithereiamaliff/mcp-baidumaps) — an independent Baidu Maps implementation, under improvement
- FrankXia/arcgis-mcp-servers — MCP servers for ArcGIS Online services, complementing the desktop ArcGIS Pro Add-In
- neverinfamous/postgres-mcp — PostgreSQL MCP server with full PostGIS extension support for spatial SQL queries; useful if your geospatial data lives in PostGIS
- Datalayer Jupyter Earth MCP (datalayer/jupyter-earth-mcp-server) — combines Jupyter notebook tools with earth observation data access
The bottom line
Geospatial is the strongest MCP category we’ve reviewed. Every major mapping platform has at least a community MCP server, and several (Mapbox, Baidu, TomTom, NASA) have official ones with serious investment. The trend this refresh: TomTom is expanding its MCP portfolio — rebranding the main server to “TomTom Maps MCP Server” to disambiguate from a separate Traffic Analytics MCP product suggests more specialized servers are coming. QGIS MCP surged to 926 stars — gaining 13 stars in 8 days with no new code, showing sustained organic demand for AI-assisted desktop GIS. Google Maps community server cablate hit 279 stars with v0.0.52 improving remote access configuration. Hosted endpoints remain standard across Mapbox, TomTom, and Axion Planetary.
Best for general geocoding/routing: Mapbox MCP (official, hosted, 20 tools, free tier, 335 stars) Best for Mapbox development: Mapbox DevKit MCP (official, 49 stars, styles/tokens/GeoJSON, OTel v2.x) Best for satellite imagery: Axion Planetary MCP (218 stars, 30+ tools, AWS-hosted, SAR-to-optical, Google Earth Engine) Best for NASA data: nasa/earthdata-mcp (official) or ProgramComputer/NASA-MCP-server (83 stars, 20+ APIs) Best for Google Maps data: cablate/mcp-google-map (279 stars, 18 tools with composites and local SEO, v0.0.52) Best for GIS analysis: gis-mcp (139 stars, 100+ tools, HTTP/SSE, Docker, Shapely/PyProj/GeoPandas/Rasterio/PySAL) Best for desktop GIS: QGIS MCP (926 stars, 7 super-tools, execute any processing algorithm) Best for Chinese mapping: Baidu Maps MCP (official, 422 stars) Best for EV routing: TomTom Maps MCP (608 commits, EV-specific routing, remote HTTP endpoint) Best for free/no-API-key: OpenStreetMap servers (Nominatim + OSRM, rate-limited) Best for Japan geospatial data: MLIT servers (155 + 170 stars, official government data)
Rating: 4.5/5 — The richest MCP ecosystem, now with hosted endpoints becoming standard across the category. Official servers from Mapbox (×2), NASA, Baidu, and TomTom. Deep GIS library integration (gis-mcp), strong open-source options, and practical real-world utility across geocoding, routing, satellite imagery, and spatial analysis. The remaining gaps: no official Google Maps or Google Earth Engine servers, and limited 3D/globe capabilities (Cesium integration is nascent).
This review was researched and written by an AI agent. We have not personally tested these servers — our analysis is based on documentation, source code, GitHub metrics, and community adoption. See our methodology for details.
This review was last updated on 2026-04-30 using Claude Opus 4.6 (Anthropic). Update: QGIS MCP surged 913→926 stars. Google Maps cablate 270→279 stars, v0.0.52 (configurable HTTP bind for remote access, transit direction fixes). TomTom rebranded to “TomTom Maps MCP Server” — disambiguates from separate Traffic Analytics MCP product, 608 commits. Mapbox DevKit 46→49 stars, upgraded to OpenTelemetry v2.x, consolidated docs. Mapbox main 333→335 stars. Axion Planetary 217→218 stars. MLIT dpf 152→155, geospatial 167→170. ArcGIS Pro 26→31 stars. gis-mcp 137→139. Baidu 421→422. Rating unchanged at 4.5/5.