NCDC Climate Data Online MCP. Pull Historical Weather Records by Asking Questions.

You're giving your AI agent direct access to NOAA's National Climatic Data Center archive. This lets you pull authoritative historical weather records—everything from daily temperature averages to yearly precipitation totals—by just asking questions about specific locations and dates.

To use this, your agent first needs to know what data exists. You can start by running list_datasets to see a list of every primary climate dataset available through NCDC, like GHCND or GSOD. Once you know the source, you'll need to figure out the scope: run list_data_categories for high-level groupings (like Temperature or Precipitation), then use list_data_types to pinpoint the specific metrics you want to track—say, Maximum Temperature or Snowfall depth.

For time granularity, checking list_data_classes shows you if you need Hourly, Daily, or Monthly summaries.

When it comes to finding a location, your agent can't just guess. It needs structure. You'll start by using list_location_categories to see general filters—City, State, Country IDs. Then, running list_locations returns a list of specific geographic points based on those parameters. If you need to find the physical infrastructure, running list_stations gives a comprehensive rundown of all active weather stations within a defined area or category.

The system also lets your agent dive deep into any single station's history by calling get_station, which retrieves full metadata and details for that specific spot. Need to know if a dataset is reliable? You can run get_dataset to get detailed information about a specific NCDC dataset, confirming exactly what it covers and how its structure works.

When you're ready to pull the numbers, your agent uses get_climate_data. This tool retrieves actual, quantified weather records—like daily maximum temperature or precipitation totals—for any specified location and date range. It handles the core query using all the data scoping established above. If you just want to know what kinds of locations are possible for filtering, running list_location_categories provides those general filters.

Everything works together like this: You first determine your location parameters using list_location_categories and then narrow down specific spots with list_locations. To build the query, you use list_data_types and list_data_classes to define what metrics and time periods matter. Once you've pinpointed a station or a general area, you can get its metadata using get_station or confirming dataset details with get_dataset.

The entire workflow funnels down into get_climate_data, which pulls the final, authoritative records for you.