Weather ChatGPT Plugins
Last reviewed
May 9, 2026
Sources
11 citations
Review status
Source-backed
Revision
v2 ยท 2,496 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
May 9, 2026
Sources
11 citations
Review status
Source-backed
Revision
v2 ยท 2,496 words
Add missing citations, update stale details, or suggest a clearer explanation.
Weather ChatGPT Plugins were a category of third party extensions for ChatGPT that supplied current conditions, forecasts, radar imagery, air quality data, and historical climate records to the chat interface. The category emerged shortly after OpenAI opened its plugin platform in March 2023 and remained active until the platform was sunset in April 2024. Weather plugins solved a persistent shortcoming of the underlying Large Language Model: a fixed knowledge cutoff that left ChatGPT unable to answer questions about present or future atmospheric conditions without an external data source.
ChatGPT plugins launched in alpha on March 23, 2023, when OpenAI announced an initial set of partners that included Expedia, FiscalNote, Instacart, Kayak, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram, and Zapier. None of those launch partners offered weather data. The system relied on a manifest file, an OpenAPI specification, and a small set of natural language descriptions that taught ChatGPT when and how to call an external API. Plugins ran only inside ChatGPT Plus, required selection of GPT-4 as the underlying model, and were limited to three active plugins per chat.
Broad rollout to all ChatGPT Plus subscribers occurred on May 12, 2023. The plugin store grew rapidly during the summer and autumn of that year, eventually crossing one thousand entries. Weather sat alongside travel, shopping, finance, and education as one of the most populated categories, partly because weather data is structured, geographically scoped, and well suited to the request and response shape that the plugin manifest required.
The category was retired together with the rest of the plugin platform. OpenAI announced Custom GPTs on November 6, 2023, opened the GPT Store on January 10, 2024, blocked new conversations with plugins on March 19, 2024, and shut down all remaining plugin sessions on April 9, 2024. Most weather plugin developers either rebuilt their integrations as Custom GPTs with Actions or moved to alternative platforms that exposed the same underlying APIs.
The pretraining corpus behind a Large Language Model ends at a fixed cutoff date. The model learns climatological averages, the names of common weather phenomena, and the structure of forecast products, but it cannot retrieve the temperature in a specific city at a specific moment. A user asking "what is the weather in Boston tomorrow" was previously answered with a refusal or a generic explanation about how forecasts work. Weather plugins closed that gap by routing such requests to a live meteorological API and returning the structured response into the conversation.
The pattern combined two strengths. The plugin supplied accurate, current data from satellites, radars, ground stations, and numerical weather prediction models. ChatGPT supplied conversational framing, multi step reasoning across the returned numbers, and natural language summaries that compared multiple locations or planning windows. Good Prompt Engineering made it possible to chain a weather lookup with a packing recommendation, an outdoor activity suggestion, or a logistics decision in a single turn.
The table below lists the most widely referenced entries in the weather category during the active life of the plugin store.
| Plugin | Developer | Launch | Coverage and features |
|---|---|---|---|
| Tomorrow.io Weather | Tomorrow.io | Late May 2023 | Hourly and fifteen day forecasts, precipitation, temperature, wind, humidity, drawing on satellite, radar, ground stations, and the Tomorrow-R1 satellite |
| Xweather | Vaisala | June 28, 2023 | Local forecasts, radar imagery, Air Quality Index, ultraviolet radiation levels, sourced from the Xweather API |
| MixerBox Weather | MixerBox | Mid 2023 | Real time conditions and forecasts for over 200,000 cities, daily forecasts up to sixteen days, three hour increments up to five days, Celsius or Fahrenheit |
| Weather Meteosource | Meteosource | 2023 | Current weather, historical data, weather statistics, air quality, and forecasts using a two layer machine learning ensemble |
| WeatherWizard | Independent | 2023 | Powered by WeatherAPI, current conditions and forecasts with temperature, wind speed, humidity, link out for detail |
| WeatherWhiz | RedAntAI | 2023 | Current, forecast, and historical weather, including chance of rain over the next minutes, hours, and up to a week |
| AeroDex | Independent (soos3d) | 2023 | METAR and TAF aviation reports from aviationweather.gov, marked for entertainment use only |
| Gimmee Air Quality | Gimmee | 2023 | Two day air quality forecast for any United States ZIP code, Air Quality Index, action day flag |
| Weather Report | Independent | 2023 | Current weather drawn from METAR aviation feeds |
Several generic plugins simply named "Weather" also appeared in the store, typically built on top of WeatherAPI or OpenWeatherMap. Their feature sets overlapped heavily and they competed primarily on response formatting and forecast horizon.
Tomorrow.io was the first dedicated weather forecast plugin to enter the ChatGPT marketplace. The Boston based weather intelligence company published the plugin in late May 2023, with reports placing the public availability between May 25 and May 31, 2023. The launch was widely covered in trade press as the first weather entry in the store.
The plugin connected ChatGPT to the Tomorrow.io API, which combines numerical weather prediction output with proprietary data including ground level weather stations, satellites, and atmospheric profiles collected by commercial aircraft. The company also operated Tomorrow-R1, a satellite in low Earth orbit at roughly 500 kilometers altitude carrying a Ka band radar instrument intended for precipitation detection. The data feed supported hourly forecasts and fifteen day outlooks, with parameters covering precipitation type and intensity, temperature, wind, humidity, cloud cover, and severe weather indicators.
Cole Swain, then vice president of product at Tomorrow.io, framed the launch as a way for users to make smarter weather based decisions and to mitigate operational and safety risk. The company highlighted ten target use cases: event planning, travel, outdoor safety, agriculture, supply chain and logistics, energy management, retail inventory, healthcare crisis preparation, emergency response, and construction scheduling.
Vaisala released the Xweather OpenAI Weather Plugin on June 28, 2023. Xweather is the brand under which the Finnish industrial measurement company Vaisala consolidates its weather, climate, and environmental services. The brand was introduced in September 2022 and unified earlier acquisitions, including AerisWeather, which Vaisala had purchased in January 2022. The plugin was the first ChatGPT entry that combined precipitation radar imagery, air quality data, and ultraviolet radiation in a single integration.
The four headline capabilities were local conditions and forecast text, near real time radar imagery for precipitation visualization, Air Quality Index values along with pollutant level discussion, and ultraviolet exposure indices for sun safety planning. Samuli Hanninen, executive vice president of Vaisala Xweather, presented the plugin as a tool that would help individuals and businesses make decisions that promote safety. Scott Mackaro, head of innovation and insights, described it as an early step toward conversational interfaces for meteorological data.
MixerBox, a Taiwan based developer that became one of the most prolific publishers in the ChatGPT plugin store, shipped MixerBox Weather alongside a portfolio that eventually included MixerBox ChatMap, MixerBox News, MixerBox Translate, MixerBox OnePlayer, and others. The weather plugin offered worldwide coverage spanning over two hundred thousand cities and exposed two main forecast modes: a daily mode that returned up to sixteen days at one day resolution, and a three hour mode that returned up to five days at three hour resolution.
Returned fields included temperature, humidity, rain probability, wind speed, atmospheric pressure, and a free text description of conditions. Users could request output in Celsius or Fahrenheit. MixerBox Weather paired naturally with MixerBox ChatMap, allowing chained prompts such as a request for the weather in Palo Alto on Saturday followed by a recommendation for picnic spots if conditions were sunny. That ability to chain plugins made MixerBox a frequent reference example in tutorial content for the plugin platform.
Weather Meteosource drew on the Meteosource API, which itself blends multiple weather models with proprietary post processing. The plugin returned current weather, forecasts, air quality, statistics, and historical records for any global location. The two layer ensemble was promoted as a way to reduce single model bias and improve accuracy on rapidly changing conditions.
WeatherWizard wrapped WeatherAPI in a small plugin that prioritized speed and simple responses. It returned temperature, wind speed, humidity, and a link to a fuller forecast page, and was popular as a low friction option for users who only wanted a quick reading without lengthy framing.
WeatherWhiz offered a wider window than most competitors, advertising current, forecast, and historical data along with chance of rain across short and longer horizons. The plugin required users to specify a location because it had no access to client geolocation.
AeroDex stood apart from the rest of the category by focusing on aviation weather. The plugin queried aviationweather.gov to return METAR and TAF reports for one or more weather stations specified by ICAO identifier. The published manifest carried an explicit disclaimer that the plugin was intended for entertainment and not for pre flight decision making, with users directed to certified weather briefings for any operational use.
Gimmee Air Quality narrowed its scope to United States air quality, returning a two day forecast for any ZIP code. The plugin reported the Air Quality Index, flagged action days, and could discuss the contributing pollutants. The output was useful for outdoor activity planning even though the plugin did not provide general meteorological data. Because air quality is closely related to outdoor weather and shares many of its planning concerns, the plugin was usually grouped with the weather category rather than with environment or health.
Weather Report focused on aviation METAR feeds for general weather, a niche position that overlapped with AeroDex on data source but targeted a more casual audience.
Alongside these named entries the store also held several plugins simply called "Weather," including a hobbyist plugin built on Replit and a Vercel deployed example called WeatherGPT. The latter was published by Vercel as a starter template, intended as an educational example of how to write a plugin manifest and serve responses from edge functions.
Most weather plugins followed the same shape. The plugin manifest declared a name, a short and a long description, an authentication method (usually no authentication for free tiers and bearer tokens for paid tiers), and a URL pointing to an OpenAPI specification. The OpenAPI document described one or more endpoints such as /forecast, /current, /airquality, or /radar, with parameters for location (commonly latitude and longitude, city name, or ZIP code), time horizon, and unit system.
When a user wrote a message that mentioned weather, ChatGPT decided whether to call the plugin based on the manifest description. If it called the plugin, it constructed a JSON request, transmitted it over HTTPS, and received a JSON response that the model summarized in natural language. The model handled common chores such as converting timezone, rounding numbers, and combining the response with a packing list or event recommendation.
Weather plugins benefited from the structure of meteorological data. A typical request was idempotent, returned a small payload, and required no authentication state, which made it easy for an LLM to use correctly. The category therefore showed unusually consistent quality compared with plugins that performed multi step transactions or required user accounts.
Weather plugins inherited the general constraints of the plugin system. Only three plugins could be active in a single conversation, so a user who wanted Tomorrow.io Weather, MixerBox ChatMap, and a calendar plugin had to leave one out. The model occasionally chose the wrong plugin or failed to call any plugin at all, especially when a question could be answered from training data. Latency was higher than a typical chat reply, and outages on the upstream provider produced confusing failure messages.
Location handling was a recurring pain point. Plugins that took city names sometimes confused similarly named cities (Springfield, Cambridge, Portland), and plugins that took latitude and longitude required ChatGPT to geocode names first, which it did with mixed accuracy. Time zone handling for forecast windows could also produce surprising results when the user and the location were in different zones.
Finally the platform offered only weak observability. Users could expand a tool call to inspect the request and response, but there was no consistent way to compare two plugins on the same query, and developers had limited insight into how the model was constructing prompts on their behalf.
When OpenAI announced Custom GPTs on November 6, 2023, the new framework subsumed plugin functionality through a feature called Actions. An Action is conceptually identical to a plugin, with the same OpenAPI based contract, but it lives inside a single GPT rather than the global plugin store. The GPT Store opened on January 10, 2024 and provided a discovery surface that replaced the plugin browser.
The announcement of plugin deprecation followed quickly. On March 19, 2024, OpenAI closed the plugin store and prevented users from starting new conversations with plugins. Existing plugin conversations remained usable until April 9, 2024, after which the platform was fully shut down. Most major weather plugin operators rebuilt their integrations as Custom GPTs. Tomorrow.io, MixerBox, and Vaisala Xweather all published GPTs that exposed the same underlying APIs through Actions, often with additional capabilities such as image generation for radar visualization or document upload for processing weather sensitive plans.
The weather category proved to be one of the most durable in the transition. Forecast data is universally useful, the underlying APIs were already designed for machine consumption, and the conversational interface advantages identified during the plugin era translated cleanly to the GPT framework.
Weather plugins were a small but representative slice of the ChatGPT plugins ecosystem. They demonstrated the value of grounding a Large Language Model in live external data, the way conversational summarization can add value on top of structured feeds, and the difficulty of building durable third party platforms on top of a chat product given how quickly OpenAI replaced plugins with Custom GPTs.
For users the weather category was one of the most consistently useful slices of the plugin store. For developers it offered a low complexity entry point into the platform.
See also: ChatGPT Plugins, ChatGPT Plugin Categories, Environment ChatGPT Plugins and Weather