WindRose
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@WindRoseWhere is it raining right now in northern Italy?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
WindRose MCP for Claude
Connect Claude to WindRose: live readings, history, climatology and records from the public WindRose network of personal weather stations and from your own stations, plus the app's models (heat-stress indices, pollutant dispersion), forecasts, the official hardware manuals, and your alerts and monitoring.
Ask things like:
Where is it raining right now in northern Italy?
Which stations within 50 km of Verona had the strongest gusts this week?
Compare October rain totals across the stations of Lombardy.
What is the all-time record low at my station, and how does this month compare with its normals?
How do I calibrate the rain collector on my Vantage Vue?
Alert me when the gust at my station goes above 60 km/h. (Business)
The server is hosted by WindRose — there is nothing to install for Claude.ai, Claude Desktop or Claude Code. This repo holds the setup guide, example configs and a tiny optional stdio bridge.
Requirements
A WindRose account on the Pro plan (read-only tools) or Business (all tools, including creating and deleting alerts and monitoring windows).
Server URL:
https://app.windroses.world/api/mcp.php(MCP Streamable HTTP).
Related MCP server: wu-weather-mcp
Claude.ai and Claude Desktop — from the connector directory (recommended)
WindRose is listed in Claude's connector directory: https://claude.ai/directory/connectors/windrose
In Claude open Settings → Connectors (or Browse connectors), search for WindRose and click Connect.
A WindRose page asks you to sign in and to allow access. That's it.
Sign-in uses OAuth 2.1 (PKCE, dynamic client registration). You can disconnect at any time from Claude, or from WindRose → Settings → API & MCP access → Connected AI apps.
Prefer adding it by hand? Settings → Connectors → Add custom connector with the URL https://app.windroses.world/api/mcp.php works the same way.
Claude Code
With sign-in (OAuth):
claude mcp add --transport http windrose https://app.windroses.world/api/mcp.php
# then inside Claude Code run /mcp and choose "Authenticate"Or with an API key (WindRose → Settings → API & MCP access → Generate key; the token is <key_id>.<key_secret>):
claude mcp add --transport http windrose https://app.windroses.world/api/mcp.php \
--header "Authorization: Bearer wr_xxxxxxxxxxxxxxxx.your_key_secret_here"A project-scoped config is in examples/mcp.json (save it as .mcp.json; it reads the key from the WINDROSE_API_KEY environment variable).
Claude Desktop with a local config file (stdio bridge)
For setups that only accept stdio servers, this repo is also a zero-dependency Node package that forwards stdio to the remote server with your API key. Add to claude_desktop_config.json (example):
{
"mcpServers": {
"windrose": {
"command": "npx",
"args": ["-y", "github:gabriele-wink/windrose-mcp-claude"],
"env": { "WINDROSE_API_KEY": "wr_xxxxxxxxxxxxxxxx.your_key_secret_here" }
}
}
}Requires Node.js 18+. Optional WINDROSE_MCP_URL points the bridge at another WindRose host.
Tools
All values are metric, with the unit in the key name (temp_c, gust_kmh, pressure_hpa, rain_mm, rain_rate_mm_h…); unix times are ts.
Tool | Access | What it answers |
| public | Latest reading of every station the user can see (all PUBLIC stations of the WindRose network + the user's own), filtered and sorted. |
| public | One number per station over a date range, from daily summaries (kept forever): sector studies across the network. |
| public | Find stations (public network + the user's own) by name/place text, area or source, with ids, position, source and last reading time. |
| public | Metadata of one station: place, coordinates, elevation, timezone, data source, what it measures now, how much history exists (daily summaries and full-resolution archive) and the retention. |
| public | Live reading of any visible station: every outdoor quantity, derived indices (heat index, wind chill, THW, WBGT, UTCI, wet bulb) and air quality; hardware/indoor values for the owner only. |
| public | Daily summaries of one station (one row per local day, up to 731 days): min/max/avg temperature with times, humidity, dew point, pressure, rain, max rain rate, mean wind, max gust with time and direction, solar, UV, evapotranspiration. |
| public | Monthly climatology of one station from its daily summaries: per month temperature means/extremes, rain total and rain days, frost/hot days, tropical nights, gusts, ET; plus normals by calendar month across years. |
| public | Records of one station over its history or a range: highest/lowest temperature with time, warmest/coldest day, wettest day and month, max rain rate, strongest gust, pressure extremes, max UV. |
| public | Time series of one station: raw archive records (max 3 days) or hourly aggregates (max 31 days): temperature, humidity, dew point, wind, gust, direction, rain, rain rate, pressure, solar, UV. |
| public | Comfort and heat-stress calculator for any conditions (real or hypothetical): dew point, heat index, wind chill, humidex, wet bulb, WBGT (Dimiceli and Liljegren) and UTCI with its stress category. |
| public | Pollutant dispersion screening (the model behind the app's plume map): EPA ISC3/SCREEN3 Gaussian plume, Cirillo-Poli below 1 m/s. |
| public | Weather forecast (Open-Meteo model, NOT a station) at any point or at a station: daily and hourly temperature, rain and probability, wind, gusts, pressure, clouds, UV, sunrise/sunset. |
| public | Historical daily weather from the ERA5 reanalysis (Open-Meteo model, NOT a station) at any point or station, up to 3 years per call: compare a station with the model, fill gaps, climate context. |
| public | Sunrise, sunset, solar noon, civil twilight and day length at a point or station on a date. |
| public | Search the official manuals of the supported hardware (Davis Vantage Pro2/Vue, WeatherLink Live/Console, AirLink; Ecowitt gateways, WS69/WS90 arrays, HP2550; Netatmo; WeatherFlow Tempest; AcuRite Atlas). |
| public | The manuals indexed for search_manual: brand, model, link, when last checked for a newer revision. |
| public | Search WindRose: stations (public network + the user's own) by name or place, and the indexed hardware manuals. |
| public | Full content of a search result: "station:" (metadata + live reading) or "manual::" (that page of the manual). |
| owner | The user's plan, its limits (stations, history, AI, exports, alerts, monitoring, API/MCP) and current usage. |
| public | Catalog of the quantities an alert rule can watch (key, label, unit family, yes/no flags). |
| owner | The user's alert rules (all stations, or one): sensor, operator, threshold (metric), cooldown, mode, enabled, last triggered. |
| write | Create a push alert rule on one of the user's stations. |
| write | Delete one of the user's alert rules by id (list_alerts gives ids). |
| write | Pause or resume one of the user's alert rules. |
| owner | The user's monitoring windows (scheduled 1-minute recordings, max 4 h each): station, name, start/end, sample count. |
| owner | Samples recorded by one of the user's monitors (metric: temp, humidity, dew point, wind, gust, pressure, rain rate, UV). |
| write | Schedule a monitoring window on one of the user's stations: the station is read every minute from starts_at to ends_at (max 4 h). |
| write | Delete one of the user's monitors and its samples. |
Access: public = the public WindRose network plus your own stations; owner = your account only; write = changes your account, Business plan only, and Claude asks before running it.
The server also exposes resources (windrose://docs/conventions, windrose://docs/capabilities, windrose://stations/mine, windrose://station/{id}) and prompts (rain_now, station_report, sector_study, compare_stations).
Data and privacy
You see every public station and your own stations. Other users' private stations are never reachable; hardware, indoor and console values are visible only to the station owner.
Forecasts and ERA5 data come from Open-Meteo models, not from stations — the tools say so.
Derived indices (heat index, WBGT, UTCI…) and the dispersion model are estimates, not certified measurements.
Station names are written by their owners and reach the model as text. Tools that change your account always need your confirmation in the client.
Troubleshooting
Symptom | Fix |
401 Missing or malformed token | Reconnect (OAuth) or check the key format |
403 MCP access is not included | MCP needs the Pro or Business plan. |
429 Rate limit exceeded | 60 requests per minute per key or connection; wait a minute. |
Write tools missing | Creating/deleting alerts and monitors needs Business, and the |
Italiano
Collega Claude a WindRose: in Claude.ai/Desktop apri Impostazioni → Connettori, cerca WindRose nella directory (https://claude.ai/directory/connectors/windrose) e premi Connetti, poi accedi con il tuo account WindRose. Serve il piano Pro (sola lettura) o Business (anche allerte e monitoraggi). In Claude Code: claude mcp add --transport http windrose https://app.windroses.world/api/mcp.php e poi /mcp.
License
MIT — see LICENSE. The WindRose service itself is subject to the WindRose terms.
This server cannot be deployed
Maintenance
Related MCP Connectors
WHOOP recovery, strain, sleep and workouts in Claude via official WHOOP OAuth. Free, open source.
- platform7nOAuthtech.p7n
Connect Claude to your Platform7n workspaces — chat, links, and tasks. One-click OAuth.
Garmin data in Claude & ChatGPT via the Garmin Health API. OAuth sign-in, no password sharing.
Pace is a remote MCP server that exposes wearable and fitness data to Claude via the Model Context Protocol. It connects to Garmin, Oura, Whoop, Polar, Fitbit and 20+ devices and provides 15 tools for querying sleep, activity, recovery, and training data. Hosted on Google Cloud Run, OAuth 2.1 authentication, Streamable HTTP transport. Instructions: First you need to create an account at: https://pacetraining.co and connect your wearables. After that you can connect the remote Server via Custom Connector in Claude and OAuth 2.1 Flow startet.
Related MCP Servers
- FlicenseBqualityDmaintenanceIntegrates the US National Weather Service API with Claude Desktop to provide real-time weather conditions, forecasts, and detailed weather metrics for any US location through natural language queries.21-
- AlicenseAqualityDmaintenanceExposes Weather Underground personal weather station data as tools in Claude, enabling queries for current conditions, daily summaries, hourly history, and station metadata.4MIT
- AlicenseNot gradedqualityDmaintenanceAn MCP server that wraps the Ambient Weather REST API. Query your personal weather stations conversationally from Claude Code, Claude.ai, or any MCP-compatible client.1MIT
- AlicenseNot gradedqualityCmaintenanceEnables Claude Desktop to query real-time weather information, forecasts, and history via the OpenWeatherMap API.6 npm98MIT