AI assistants are good at reasoning but blind to the present. Ask Claude about the current fire risk near a property and it can only answer from training data that ended months ago. The Model Context Protocol (MCP) fixes this: it is an open standard that lets AI assistants call live tools. PixelGust now runs a public MCP server, so any compatible assistant can pull real weather, climate, terrain, and hazard data for any point on Earth, mid-conversation.
Server URL: https://pixelgust.com/mcp (Streamable HTTP). Works with no signup at free-tier limits. Add it to Claude, Cursor, or any MCP client and start asking location questions.
What You Can Ask #
Once connected, your assistant answers questions like these with live data instead of guesses:
- "What is the fire risk in Athens versus Marseille right now?"
- "Has this vineyard's location gotten drier over the last decade?"
- "Compare the terrain, flood susceptibility, and soil of these two coordinates."
- "What weather should I expect at 46.5, 11.3 in early September?"
- "What is the average slope and population inside this polygon?"
A real exchange from today, with the assistant calling the server twice and comparing:
| Athens | Marseille | |
|---|---|---|
| Fire Weather Index | 34.1 (High) | 30.5 (High) |
| Temperature | 31.6°C | 26.6°C |
| Relative humidity | 42% | 60% |
| Wind | 3.5 m/s SW | 5.1 m/s S |
The assistant concluded that both cities sit in the High danger class, with Athens closer to Very High because heat and dryness outweigh Marseille's stronger wind. That is the kind of grounded, current answer MCP makes possible.
The Nine Tools #
get_current_weather: temperature, humidity, wind, precipitation, pressure, cloud cover, and solar radiation for a point (NOAA GFS, updated every 6 hours)get_weather_forecast: daily forecast up to 7 days ahead** get_historical_climate**: 10-year climate averages (2015-2025, ERA5), annual or by month** get_climate_timeseries**: monthly values 2015-2025 for one variable, for trend analysis** get_terrain**: elevation, slope, and aspect at 30m (Copernicus DEM), plus wetness index** get_hazards**: Fire Weather Index, soil erosion risk (RUSLE), flood susceptibility (TWI)** get_environment**: NDVI vegetation index, evapotranspiration, land cover class** get_proximity**: distance to the nearest city, airport, port, hospital, and power plant** get_polygon_stats**: aggregate statistics over a custom polygon area
How to Connect #
Claude Code
claude mcp add --transport http pixelgust https://pixelgust.com/mcp
Claude Desktop and claude.ai
Settings → Connectors → Add custom connector → paste https://pixelgust.com/mcp
.
Cursor and other MCP clients
Add a Streamable HTTP server with the same URL in your client's MCP configuration.
Higher limits with an API key
Keyless access uses free-tier limits (50 calls per day). For more, create a free account at pixelgust.com/app, generate an API key from the API Keys menu, and pass it as a header:
claude mcp add --transport http pixelgust https://pixelgust.com/mcp \
--header "X-API-Key: pk_your_key"
The key's plan applies automatically: paid tiers raise the daily quota and unlock larger polygon areas. See the API documentation for details.
Give Your Assistant Real Environmental Data
One URL. Nine tools. Any location on Earth. Free to start.
Why This Matters #
Agents are becoming the way professionals interact with data. An agronomist asking an assistant to compare drought stress across fields, an analyst screening properties for climate risk, or a researcher pulling decade-long climate series should not have to open a GIS suite or write API integration code first. With MCP, the assistant does the calling, and the human does the asking.
Under the hood the server exposes the same data as the PixelGust REST API: NOAA GFS nowcasts and forecasts, ERA5 reanalysis, Copernicus DEM terrain, MODIS evapotranspiration, ESA WorldCover, SoilGrids, and the Canadian Fire Weather Index computed daily. Same sources, new interface.