aggregate_weather_data
ChatGPTCompute one statistic — max, min, avg, sum, or count — server-side over the hourly WeatherOptics forecast at one or many locations, for either a weather variable (temperature, wind_gust, total_precipitation, snow_flag, …) or a 0-10 impact index (road, flood, life_property, disruption, power, wildfire_spread). Answers questions such as 'total rain at the Chicago yard over the next 48 hours' (sum of total_precipitation), 'peak road risk across my terminals this week' (max of road), or 'how many hours below freezing with precipitation at Denver' (count of total_precipitation with filter temperature lt 32). An optional filter restricts the statistic to the hours where a field (the same one or another) meets a threshold. Returns one value per location with at (local time of the max/min hour), hours_considered, hours_matching_filter, and — for impact indices — the canonical severity band plus the fixed official blurb for that value; and a cross-location summary (e.g. the fleet-wide maximum and which location holds it). The window is 1-168 hours from now; past periods come from get_daily_historical. A location with error has no data — its value is unknown, not zero; failures are itemized in warning and the call is marked as an error. For the full hourly series, or a chart, use get_weather_series or get_impact_series. For the standard peak-risk-and-when summary across all indices, use get_impact_forecast. For a multi-variable weather digest with daily highs/lows, use get_weather_forecast. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
find_coordinates_for_place
ChatGPTGeocoding helper. Given a place name, address, city, or landmark (e.g. 'Richmond, VA' or '1600 Pennsylvania Ave'), return candidate {name, latitude, longitude} matches from the WeatherOptics geocoding service. Use this before calling the forecast tools when you only have a free-text place description and don't already have reliable coordinates — for city-scale questions, city-center coordinates from your own knowledge are sufficient; this tool is for street addresses and ambiguous names. Pick the best match and pass its latitude/longitude (and the user's asset name as label) into the forecast tool. If multiple matches look plausible, ask the user to confirm rather than guessing. The forecast tools also accept a place field directly and resolve it with this same service — this standalone tool is for exploring candidate matches first. It looks up arbitrary places — to search records the user has saved in WeatherOptics, use search_locations. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_active_cyclones
ChatGPTUSE THIS TOOL for broad tropical questions with no specific place: 'any hurricane activity?', 'what tropical storms are active?', 'anything brewing in the Gulf / Atlantic?', 'update on typhoons'. Lists every active tropical cyclone worldwide from the WeatherOptics cyclone feed (NHC: Atlantic, Eastern and Central Pacific; JTWC: Western Pacific and other basins), each with its current position, wind, gust and category and the full official forecast track; track_ends_at is the forecast horizon (the last official position), and the forecast covers exactly that span. Also lists tropical disturbances — areas that may develop — with the forecaster's formation chance over 48 hours and 7 days and the outlook text, verbatim. Takes no input. For storms, alerts or warnings near a specific place use get_severe_events_near; for the 0-10 impact outlook at a location use get_impact_forecast; to show storms on a map use get_tile_map. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_daily_historical
ChatGPTDay-by-day WeatherOptics historical summaries at one location for up to 366 local days, for questions like 'how much rain did Seattle get last month?', 'what was the coldest night this winter?' or 'what was road risk at the depot during the February storm?'. Weather days use the same fields and are computed the same way as the daily summaries in get_weather_forecast (temperature_high, temperature_low, total_precipitation, total_rain_precipitation, max_wind_gust, ...); impact indices give each day's peak (<index>_peak). summary gives window extremes with their dates and sums of the daily totals. WeatherOptics historical data covers North America and Europe from 2022-05-04 and elsewhere from 2022-12-02 (API docs); new data typically arrives within 6 hours, so the most recent hours may be missing (data_ends_at shows where the data ends). Impact power is the current Power Outage Index V2. Long windows are fetched as several one-week WeatherOptics requests, reported in upstream_calls. Dates with no data are listed in missing_dates; a day with only part of its hours has hours_available. For hour-by-hour detail within 7 days, get_hourly_historical. Pass a label for the location; it is echoed back verbatim. A result with error has no data — missing data is unknown, not calm or dry conditions. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_hourly_historical
ChatGPTHourly WeatherOptics historical data at one location for up to 7 local days: weather variables (temperature, wind, gusts, precipitation, etc.) and/or 0-10 impact indices (road, flood, disruption, life_property, power). For questions like 'when did it rain hardest on Tuesday?' or 'how windy was it at the warehouse last night?'. WeatherOptics historical data covers North America and Europe from 2022-05-04 and elsewhere from 2022-12-02 (API docs); new data typically arrives within 6 hours, so the most recent hours may be missing (data_ends_at shows where the data ends). Impact power is the current Power Outage Index V2. Each call makes one WeatherOptics request per data type requested (weather, impact). For longer windows or day-by-day summaries, get_daily_historical covers up to a year. For future conditions, get_weather_forecast or get_impact_forecast; for a past drive along a route, get_rightroute_forecast with a past departure_time. Pass a label for the location; it is echoed back verbatim. A result with error has no data — missing data is unknown, not calm or dry conditions. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_impact_forecast
ChatGPTAssess weather risk for one or many places over the next 7 days. Use this whenever the user has a location, address, or a list/file of sites, assets, stores, or warehouses and wants to know which are at risk, when bad weather hits, or how severe it will be. Also answers 'is there wildfire spread risk at/near X?' — a 0-10 wildfire_spread score at the point (opt-in index, with the active fire's name). Returns the peak WeatherOptics index per location and when it occurs. All indices are a 0-10 scale: flood = flooding risk, road = driving/road-conditions risk, life_property = danger to life & property, disruption = business disruption, power = power-outage risk. One opt-in wildfire index (not in the default set; pass it in indices): wildfire_spread = the risk that an ACTIVE fire's spread reaches this point, with the fire's name in wildfire_name (a peak of 0 with no name = no active fire's spread reaches the point) — e.g. 'is there wildfire spread risk at X?' is indices 'wildfire_spread'. This tool covers the forecast only. Past dates come from get_hourly_historical (hourly, up to 7 days) and get_daily_historical (day by day, up to a year). Closures and road surface conditions being reported right now come from get_live_road_status. Pass a label for every location using the name/ID from the user's data (e.g. the asset name or row id); results are keyed by these returned labels and coordinates. Severity bands (low/moderate/high/extreme) are WeatherOptics-canonical: the returned band and overall values are authoritative, not re-derivable from the numbers. Each peak's blurb is fixed official WeatherOptics wording for that score — reproduce it verbatim, not paraphrased. RELIABILITY: requested/succeeded/failed report coverage. A failed location has NO data — its risk is unknown, not low; failures are itemized in warning and the call is marked as an error. If the user asks to CHART / PLOT / GRAPH / VISUALIZE / SEE risk over time at a single location, use get_impact_series instead — this tool returns only a peak-and-when summary, not the full hourly series. For plain weather amounts (temperature, inches of rain/snow, wind speed), use get_weather_forecast instead. If the user asks about HURRICANES, tropical storms, tornadoes, hail, active warnings, or named storm systems near a location, use get_severe_events_near instead — this tool only returns generic 0-10 risk indices and knows nothing about specific storms. Every active tropical cyclone and disturbance worldwide, with no location needed, is listed by get_active_cyclones. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_impact_series
ChatGPTUSE THIS TOOL when the user asks to CHART, PLOT, GRAPH, VISUALIZE, DRAW, or SEE a WeatherOptics risk over time — or otherwise wants the raw hourly risk series. Do NOT use get_impact_forecast for these requests — that tool returns only a peak-and-when summary, not the full hourly series. This tool returns points[] (one entry per hour with time_iso, local_time_iso, and value) for a single location and a single index (road, flood, life_property, disruption, power, or wildfire_spread — active-fire spread risk, 144-hour horizon, Wildfire risk permission) over the forecast window. If your chat client can render charts natively, chart the points yourself. A chart_url is also returned for MCP clients that can embed image URLs — use it only as a fallback. Pass a label for the location; it is echoed back verbatim. For a multi-location or summary risk question (no chart), use get_impact_forecast. For a plain weather-variable series, use get_weather_series. For a spatial map of an index over an area (not a graph), use get_tile_map. Past dates come from get_hourly_historical. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_live_road_status
ChatGPTUSE THIS TOOL for questions about what is being REPORTED on the roads right now: 'is I-80 closed in Wyoming?', 'are roads icy around Denver?', 'any closures or accidents on my route?', 'truck restrictions near Flagstaff?'. Wraps the WeatherOptics Live Road Status API: state and provincial 511 / DOT reports, collected about once an hour — each item's last_updated is the source's own time, so reports can lag events on the ground. Returns road closures, road surface status (normalized surface_condition: dry_normal, wet_slush (wet pavement as well as slush — the item's description says which), black_ice, snowy, low_visibility, severe_driving_conditions, road_closed, crosswind_danger, unknown), traffic incidents, truck warnings and weather events by default; road_work, traffic_congestion and special_event on request via types. Search EITHER around location (radius_miles, default 25) OR along a route: pass the route_polyline array from get_rightroute_forecast (called with include_route_polyline: true) UNCHANGED, and radius_miles is the corridor half-width (default 5); route items carry miles_along_route and come back in route order. Dry/normal surface reports are counted in surface_conditions and their roadways listed in normal_surface_roadways rather than itemized; at most 50 items are returned (a warning says when more matched). coverage_gaps names states in the searched area whose 511 sources do not provide a requested type (e.g. no surface status in Florida) — an empty list there means not collected, not clear. Coverage is US-wide; Canadian reports need a permission this connection may not have (the result says so). This is reported current conditions, not a forecast: for expected weather slowdowns and hazards on a future drive use get_rightroute_forecast; for the 0-10 forecast road-risk index at a place use get_impact_forecast; for severe storms and NWS warnings use get_severe_events_near. This tool returns no map, and get_tile_map has no road status layer. Each item's nearest_point is the single point closest to the search, not the extent of the closure or condition, so plotting those points does not show where a report starts and ends; the WeatherOptics portal map draws each report's full extent (Map Layers → Traffic Layers). Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_rightroute_forecast
ChatGPTUSE THIS TOOL for questions like 'if I'm driving from X to Y at time T, what weather delays should I expect?', 'is my drive going to be bad?', or retrospectives like 'what were conditions on my route last Tuesday?'. Wraps the WeatherOptics RightRoute Forecast API — combines the routing engine with the WeatherOptics impact forecast to return weather-adjusted travel time, adjusted arrival time, and expected slowdown percentage, plus (by default) a compact per-route narrative of the weather + impact conditions along the way. A past departure_time automatically switches to the historical reconstruction and returns mode: "historical". Historical routes cover the contiguous US for departures from 2007-01-01, use less granular weather data than forecast routes, gain new data every 24 hours, and do not include wildfire risk, government weather alerts or power outage risk (the returned note says so). For past weather or risk at a single place (not a drive), get_hourly_historical and get_daily_historical cover it. Origin and destination take latitude/longitude directly, or a free-text place name resolved server-side via WeatherOptics geocoding (the chosen match is echoed in each point's resolved_place). City-center precision is sufficient — weather hazards vary at the tens-of-miles scale; for an ambiguous name (e.g. 'Springfield' with no state), a more specific place string or explicit coordinates avoids a wrong match. Pass label on each point for narration; echoed back verbatim. Result semantics: hazard_callouts lists every elevated risk along the route (wildfire spread, flood, road, etc.) with canonical blurbs (official fixed wording) — none is optional context; each represents real risk on the trip. Each callout's where array gives the contiguous route segments (start/end time + coordinates) where that hazard is elevated. A 0% slowdown does NOT mean no weather risk: slowdown only measures drive time, and e.g. an active wildfire near the route shows up only in hazard_callouts. overall_hazard_level is the canonical severity of the whole trip (computed server-side from the worst callout) — it, not the slowdown percentage, is the trip's bottom line. An elevated wildfire_spread callout means an ACTIVE fire near the route. official_guidance (present when the hazard level is moderate or above) names official sources the user should also consult, and presentation_guidance carries per-result reporting notes for these fields. For severe storms or warnings near a single point (not a route), use get_severe_events_near. For closures, incidents and road surface conditions being reported on the route right now, pass its route_polyline to get_live_road_status. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_severe_events_near
ChatGPTUSE THIS TOOL for questions like 'what severe storms are around Chicago?', 'will it hail on me?', 'is there a tornado warning?', 'is a hurricane heading for Tampa?', 'any wildfires near me?' (incidents and alerts only — it has no 0-10 spread score; 'is there wildfire spread risk at X?' is get_impact_forecast with indices 'wildfire_spread'), or anything asking about active severe weather / warnings / wildfire incidents near a location. Wraps the WeatherOptics events feed (SevereCast for next-hour hail/tornado/damaging-wind cells, storms + cyclones for hurricanes/winter storms days out, us_wildfires for active US wildfire incidents with name/size/perimeter) plus NWS alerts pass-through for official tornado warnings, severe thunderstorm warnings, etc. Returns events normalized to {source, event_type, severity, headline, starts_at, ends_at, distance_miles, properties}, sorted closest first — at most the 50 closest; a warning says when more matched. SevereCast is CONUS-only. Pass a label for the location; it is echoed back verbatim. The location takes latitude/longitude or a free-text place name resolved server-side (echoed in resolved_place); city-center precision is sufficient — severe-weather queries are regional, so a more precise address is not needed. Each call covers one location: events outside the searched area are not returned, so calls at guessed locations silently miss events elsewhere. Tropical cyclones and disturbances anywhere in the world ('any hurricane activity?', 'anything brewing in the Gulf?') are all listed by get_active_cyclones, with no location needed; for any other broad question with no place ('what severe weather is happening?'), the location has to come from the user. For a generic 0-10 risk outlook over the coming days (not specific active storms/warnings), use get_impact_forecast. For the numeric wildfire risk score at a point ('is there wildfire spread risk at X?'), use get_impact_forecast with indices 'wildfire_spread' — this tool lists incidents, not a per-point score. To show storms or radar on a map, use get_tile_map (layer 'radar'); for fire perimeters and spread polygons on a map, layer 'wildfire'. For road closures, accidents and icy/snowy road reports, use get_live_road_status. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_tile_map
ChatGPTRender a map image of a WeatherOptics data layer (road risk, flood, wind, radar, wildfire, etc.) centered on a location — for prompts like 'show me the road-risk map around Chicago'. Layer 'wildfire' draws active fire perimeters and their spread polygons over the basemap (no tiles); the fires in view are summarized in fires (acres, percent contained, highest spread index) and named in note. Can also draw a ROUTE map: for 'map my route' / 'weather-aware route map' requests, first call get_rightroute_forecast with include_route_polyline: true, then pass its route_polyline here — renders basemap + the route hazard-colored by drive-time conditions (no tile layer; see route_polyline). In clients that support MCP Apps the map renders inline automatically via the attached UI template; the map image and tile URLs go only to that widget (they embed the caller's key), so this result carries no link to offer. For named cities, city-center coordinates from your own knowledge are sufficient — a street address is not needed. For a time-series graph at a single point (not a map), use get_impact_series or get_weather_series. For the numeric wildfire risk score at a point ('is there wildfire spread risk at X?'), use get_impact_forecast with indices 'wildfire_spread' — a map layer carries no per-point 0-10 score. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_weather_forecast
ChatGPTGet plain weather variables (temperature, precipitation, wind, humidity, heat index, etc.) at one or many places, from the WeatherOptics blended forecast model. Hourly rows cover the next 7 days (168 hours), the window where hourly detail is reliable. The daily summaries run past that, typically about 16 days including today; days beyond 7 are best used for trends (warmer or colder, wetter or drier, a windy stretch), since exact amounts and timing are less reliable that far out. Use this for location-specific non-risk questions like 'will it rain in Chicago Thursday?', 'how hot next week?', or 'what's the wind forecast for my warehouses?'. This tool covers the forecast only. Past dates come from get_hourly_historical (hourly, up to 7 days) and get_daily_historical (day by day, up to a year). For weather risk questions (road danger, flooding, business disruption, life/property, power outage), use get_impact_forecast instead — that returns 0-10 canonical risk scores. Each location returns both an hourly series (filtered to the requested variables) and a compact daily summary. Pass a label for every location using the name/ID from the user's data; results are keyed by these returned labels. A location with error has NO data — missing data is unknown, not clear/fine conditions. If the user asks to CHART / PLOT / GRAPH / VISUALIZE / SEE a weather variable over time at a single location, use get_weather_series instead — that tool returns a rendered chart image plus the same underlying points. For hail, tornadoes, hurricanes, or active warnings near a location, use get_severe_events_near; every active tropical cyclone and disturbance worldwide is listed by get_active_cyclones. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
get_weather_series
ChatGPTUSE THIS TOOL when the user asks to CHART, PLOT, GRAPH, VISUALIZE, DRAW, or SEE a single weather variable over time — or otherwise wants the raw hourly series for one variable. Returns points[] (one entry per hour with time_iso, local_time_iso, and value) for a single location and a single variable (temperature, precipitation, wind, humidity, etc.). For a multi-variable weather digest, use get_weather_forecast instead. If your chat client can render charts natively, chart the points yourself. A chart_url is also returned as a fallback for MCP clients that can embed image URLs. Pass a label for the location; it is echoed back verbatim. For a risk-index series (road, flood, etc.), use get_impact_series. For a spatial map of a layer over an area (not a graph), use get_tile_map. Past dates come from get_hourly_historical. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
search_locations
ChatGPTSearch the user's own saved locations (their "locations" list in WeatherOptics). Use this whenever the user asks about their places — e.g. "my Chicago locations", "my locations in Texas", or just "my locations". Each record has latitude/longitude ready to feed into get_impact_forecast (label = the record's name). If the result count exceeds what's useful, pass a more specific query. This searches only records the user has saved in WeatherOptics — to find coordinates for an arbitrary place name, use find_coordinates_for_place. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
search_routes
ChatGPTSearches the user's own saved WeatherOptics routes — planned trips, loads/shipments, and recurring lanes — by route name, origin/destination label, or external ID (e.g. LOAD-1001, R-1042), and returns each match's id, name, origin and destination (label plus latitude/longitude), waypoints, departure_time, recurring-lane cron, status, and assigned vehicle. Typical questions: "my routes", "my Dallas to Memphis lane", "route R-1042", "my load LOAD-1001", "which of my routes leave tomorrow?". When the user names a saved route, load, or lane (by name or ID), it is looked up here first so the forecast uses the saved origin, destination, waypoints, and departure_time. Results contain no weather or delay data; a forecast along one of these routes comes from get_rightroute_forecast given the route's origin/destination coordinates and departure_time. Saved point locations (warehouses, stores) are searched with search_locations, vehicles with search_vehicles, and arbitrary place names are looked up with find_coordinates_for_place. Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
search_vehicles
ChatGPTSearch the user's own vehicles / fleet assets. Use this whenever the user asks about their trucks, vans, or fleet — e.g. "my fleet", "my Dallas trucks", or "vehicle V-103". Each record has latitude/longitude (the vehicle's current position) ready to feed into get_impact_forecast (label = vehicle name or external_id). Note: results are a point-in-time snapshot; conditions evolve, and live data is in the WeatherOptics portal (https://portal.weatheroptics.co).
aggregate_weather_data
Claudefind_coordinates_for_place
Claudeget_active_cyclones
Claudeget_daily_historical
Claudeget_hourly_historical
Claudeget_impact_forecast
Claudeget_impact_series
Claudeget_live_road_status
Claudeget_rightroute_forecast
Claudeget_severe_events_near
Claudeget_tile_map
Claudeget_weather_forecast
Claudeget_weather_series
Claudesearch_locations
Claudesearch_routes
Claudesearch_vehicles
Claude