io.github.cyanheads/noaa-climate-mcp-server

Fetch historical observation records from a NOAA CDO dataset for a given date range. Requires datasetId, startDate, and endDate; optionally filter stations, locations, and datatypes. The encoded query is limited to 4000 characters; reduce filters and make separate calls if exceeded. Sub-daily, daily, and radar datasets allow 1 year; GSOM/GSOY/NORMAL_MLY/NORMAL_ANN allow 10 years, through the end of the corresponding calendar month. For NORMAL_* use the API proxy year 2010. Request units=metric or units=standard: omitted units retain datatype-specific raw scales. Examples: GHCND TMAX/TMIN tenths °C, PRCP tenths mm, SNOW/SNWD whole mm; GSOM/GSOY temperature/precipitation °C/mm; NORMAL_* temperature normals tenths °F and precipitation normals hundredths inches; PRECIP_HLY/HPCP and PRECIP_15/QPCP hundredths inches. These examples are not universal conversion rules. Missing precipitation rows omit value and carry missing=true; GHCND flags expose quality-check failures without dropping values.

Search for geographic locations by category (CITY, ST, CNTY, CNTRY, ZIP, CLIM_REG, etc.). Returns location IDs used in station search and data queries. Without locationCategoryId, returns all location types; noaa_climate_list_location_categories lists the valid values. Use locationCategoryId=ST to list US states (51 entries — small enough to retrieve completely). To find a location by name, pass nameContains alongside locationCategoryId — the CDO API has no name parameter, so this server enumerates the category and matches the substring itself. It works for any category under the size limit stated on nameContains, which is every category except ZIP; a datasetId or datacategoryId filter can bring a category back under that limit. For a category still too large, sort alphabetically with sortField=name and page through results. Location IDs: states as FIPS:37 (NC), cities as CITY:US530018 (Seattle), zip codes as ZIP:98101, countries as FIPS:US. Obtain location IDs here, then pass them to noaa_climate_find_stations or noaa_climate_fetch_data.

Search for weather observation stations by location, bounding box, dataset, and data type. Returns station IDs, names, coordinates, elevation, and data coverage dates. Filter by locationId (e.g., "FIPS:37" for all NC stations), extent (lat/lon bounding box), datasetId, datatypeId, and date range. Station IDs returned here are used as stationId in noaa_climate_fetch_data. A station must have data for the dataset and date range you want — filter by datasetId and startDate/endDate to ensure compatibility. Common station ID formats: GHCND:USW00024233, COOP:010008.

Query NOAA/NCEI’s retired Billion-Dollar Weather and Climate Disasters archive (1980–2024), with CPI-adjusted and unadjusted costs and deaths. No years after 2024 are added; CPI adjustments are preserved from the archive, not rebased to current prices. Every cost is in WHOLE US DOLLARS, converted from each export’s declared unit (declaredCostUnit): millions per event, billions per national year, millions per state year. Default calls return individual disasters; summary=true returns per-year counts and costs by class plus an "All Disasters" total. Filter with startYear/endYear (overlapping disasters included), disasterType (exact NCEI class, e.g. "Tropical Cyclone"), minCostInUsd, and state (two-letter postal code). coveredYears reports the span actually present in the export. Under a state scope, per-event rows carry the NATIONAL cost of disasters that reached that state, never a state share, so summing states double-counts; per-year rows carry a binned range instead of a point estimate. Separate from CDO and noaa_climate_search_storm_events; no token required.

Fetch full metadata for a single weather station by its ID (e.g., "GHCND:USW00024233", "COOP:010008"). Returns name, coordinates, elevation, and the full date range for which data is available. Use when you have a station ID from noaa_climate_find_stations and want its complete details, or to verify a station before querying data.

List data categories that group related data types — Temperature, Precipitation, Wind, Pressure, Sunshine, Sky cover, Weather Type, and more. Use to discover what types of measurements are available before calling noaa_climate_list_data_types. Optionally filter by dataset, location, station, or date range. There are 42 categories in total.

List available data types (measurement labels like TMAX, TMIN, PRCP, SNOW) for a given dataset or category. Pass a datasetId to see what is measured in that dataset, or a datacategoryId (e.g., "TEMP") to see all temperature-related types. Hundreds of types exist across all datasets. Use this before calling noaa_climate_fetch_data when the data type IDs are unknown. Common GHCND types: TMAX (max temperature), TMIN (min temperature), PRCP (precipitation), SNOW (snowfall), SNWD (snow depth), AWND (average wind speed).

List available NOAA CDO datasets with their IDs, names, and temporal coverage. Returns all ~11 datasets by default (no required parameters). Optionally filter to datasets that contain a specific data type, cover a location or station, or overlap a date range. Common datasets: GHCND (daily observations, 1763–present), GSOM (monthly summaries), GSOY (annual summaries), NORMAL_DLY/MLY/ANN/HLY (1981–2010 climate normals). Use this first to discover available datasets before calling noaa_climate_fetch_data.

List the location categories that scope noaa_climate_find_locations — CITY, CLIM_DIV, CLIM_REG, CNTRY, CNTY, HYD_ACC, HYD_CAT, HYD_REG, HYD_SUB, ST, US_TERR, and ZIP. Call this first when you do not know which locationCategoryId to pass, including the hydrological categories needed to reach a basin or region. There are 12 categories in total. This endpoint takes pagination and sort only — NOAA CDO ignores dataset, location, station, and date filters here, so none are offered.

Search the NCEI Storm Events Database for one calendar year — tornadoes, hail, floods, hurricanes, winter storms, heat, and every other NWS Storm Data event type, with magnitude, direct and indirect casualties, property and crop damage, and narratives. This corpus carries discrete severe-weather events rather than CDO station observations, needs no token, and publishes one bulk file per year, so year is required. Filter by episodeId from a returned event, state (full NCEI name such as "FLORIDA", not "FL"), eventType (exact NWS label such as "Tornado", "Flash Flood", or "Hurricane (Typhoon)", case-insensitive), begin month, inclusive local-calendar startDate/endDate overlap, and minDamageInUsd. Date bounds exclude missing, invalid, or reversed source spans before damage accounting, totals, and paging; cross-year bounds still search only the selected year export. Damage retains its raw magnitude-suffixed cell and parsed dollar amount; unreported figures are omitted, and minDamageInUsd discloses how many unknown figures it excluded. Events reference distinct page-local episodeNarratives by episodeNarrativeRef; eventNarrative stays on each event, and rendered narratives preserve literal source text. Optional local ISO beginDate/endDate values can feed CDO date filters; raw NCEI date-time and timezone remain available. Source row order, limit/offset pagination, and filtered totalCount are preserved; appliedDateRange echoes supplied bounds.