WeatherWRX MCP

Connect AI assistants to your WeatherWRX weather context

WeatherWRX exposes a hosted Model Context Protocol server so approved AI clients can read scoped forecast, alert, calendar, and briefing context through OAuth instead of copying private weather setup into a chat.

Open WeatherWRX MCP endpoint

Protected resource

https://weatherwrx.com/mcp

Transport: Streamable HTTP
OAuth: Authorization code + PKCE S256
Manage access: Settings > Connected Apps

MCP-only subscription

Agent MCP is $10.00/month with no free trial.

An AI agent can discover the product and purchase URL from /.well-known/weatherwrx-agent.json. The agent then hands checkout to the account owner, who signs in and authorizes the charge. Agents cannot charge a card without that approval.

Subscribe to Agent MCP

Forecast and alert context

Ask questions across selected WeatherWRX locations using cache-backed forecasts, government alerts, formula incidents, and dashboard briefings.

Calendar-aware weather

Let approved clients use Calendar Cabinet subscriptions and sanitized primary-calendar availability without exposing raw event text or calendar IDs.

Scoped actions

Use OAuth scopes to keep read-only tools separate from optional write tools for locations, alert formulas, and Calendar Cabinet subscriptions.

BYOK in Ask WeatherWRX

Bring your own OpenAI, Claude, or Gemini key in WeatherWRX so Ask WeatherWRX can use your existing provider account to reason over your forecasts and alerts.

Before connecting

Choose the access path that matches your assistant

Ask WeatherWRX BYOK is the simplest path when you want WeatherWRX itself to call your existing OpenAI, Claude, or Gemini provider account. External assistants use the MCP endpoint below and complete OAuth before WeatherWRX returns any user-scoped data.

Ask WeatherWRX BYOK

Go to WeatherWRX settings, choose your LLM provider, and save your provider API key. WeatherWRX encrypts the key and uses it only for Ask WeatherWRX requests you make.

External MCP clients

Add the MCP endpoint in the assistant or CLI, complete WeatherWRX OAuth, and approve the locations, subscriptions, and scopes the assistant can use.

Revoke anytime

Review or revoke connected apps from Settings > Connected Apps. A revoked grant can no longer use WeatherWRX MCP tools.

Setup instructions

Connect OpenAI, Claude, Gemini, or your own client

Use these setup screens as a field guide. Provider labels can change, but the WeatherWRX endpoint, OAuth metadata, and scope model are the same across MCP clients.

OpenAI

Use ChatGPT custom connectors or the OpenAI Responses API remote MCP tool. ChatGPT handles the OAuth authorization-code flow when the connector is created and used.

1In ChatGPT, open Settings, then Apps & Connectors. Enable developer mode if your account requires it.
2Create a connector named WeatherWRX and use the WeatherWRX MCP endpoint as the connector URL.
3When ChatGPT prompts for authorization, sign in to WeatherWRX, review the requested scopes, and approve only the access you want that connector to have.
4Start a new chat, add the WeatherWRX connector from the tools menu, and ask a weather question tied to your selected locations.

OpenAI: Connect from ChatGPT OpenAI: MCP authentication

ChatGPT connector setup

Connector name

WeatherWRX

Connector URL

https://weatherwrx.com/mcp

Auth

OAuth 2.1, PKCE S256

After Create, ChatGPT lists the WeatherWRX tools it can use.

{
  "type": "mcp",
  "server_label": "weatherwrx",
  "server_url": "https://weatherwrx.com/mcp",
  "require_approval": "always"
}

Claude

Use Claude custom connectors for chat experiences, or the Claude Messages API MCP connector when your application manages OAuth tokens separately.

1Open Claude settings and add a custom connector for a remote MCP server.
2Set the connector URL to the WeatherWRX MCP endpoint and continue to the OAuth sign-in step.
3Approve the WeatherWRX scopes that match the work you want Claude to do.
4For API usage, pass WeatherWRX as an MCP server and include the OAuth bearer token your app obtained for the current user.

Claude: Custom connectors Claude API: MCP connector

Claude custom connector

Connector

WeatherWRX

Remote MCP URL

https://weatherwrx.com/mcp

Permissions

Review each tool approval

Disable write-capable tools when you only need read-only weather answers.

{
  "type": "url",
  "url": "https://weatherwrx.com/mcp",
  "name": "weatherwrx",
  "authorization_token": "WEATHERWRX_OAUTH_ACCESS_TOKEN"
}

Gemini

Gemini CLI can connect to remote MCP servers with Streamable HTTP and can discover OAuth metadata from the server when authentication is required.

1Install and sign in to Gemini CLI with the Google account or Gemini API key you normally use.
2Add WeatherWRX under mcpServers in your Gemini settings file.
3Run Gemini CLI and use /mcp auth weatherwrx if the OAuth browser flow does not start automatically.
4Ask for a weather briefing and confirm any tool call that requests access to WeatherWRX data.

Gemini CLI: MCP server integration

Gemini CLI MCP settings

Server key

weatherwrx

httpUrl

https://weatherwrx.com/mcp

OAuth

dynamic discovery

Gemini stores OAuth tokens locally and refreshes them when possible.

{
  "mcpServers": {
    "weatherwrx": {
      "httpUrl": "https://weatherwrx.com/mcp",
      "timeout": 30000,
      "trust": false
    }
  }
}

Your own LLM

Any MCP-capable client can connect if it supports Streamable HTTP, OAuth discovery, authorization-code with PKCE, resource indicators, and bearer tokens.

1Send an unauthenticated request to the MCP endpoint and read the WWW-Authenticate challenge.
2Fetch the protected resource metadata, then fetch the WeatherWRX OAuth authorization-server metadata.
3Register dynamically when your client supports DCR, or use the client identity flow your MCP host provides.
4Run authorization-code with PKCE S256, include the resource parameter for the MCP endpoint, and send the resulting bearer token on every MCP request.

MCP authorization specification

Custom client OAuth checklist

Resource

https://weatherwrx.com/mcp

Auth metadata

https://weatherwrx.com/.well-known/oauth-authorization-server

Token header

Authorization: Bearer <token>

Reject tokens whose audience is not the WeatherWRX MCP resource.

Automation examples

Schedule morning WeatherWRX reports from your LLM

Use these patterns when you want an assistant to send a daily weather briefing before plans start. Each example keeps WeatherWRX data behind OAuth, uses read-first scopes, and asks the model to omit sensitive location and calendar details.

ChatGPT morning task

Personal daily briefing in ChatGPT

Best for: Personal daily briefing in ChatGPT
Schedule: Weekdays at 7:00 AM

Starter prompt

Every weekday at 7:00 AM, use the WeatherWRX connector to list my WeatherWRX locations, get my dashboard briefing, and send a concise morning weather report. Start with active alerts and high-risk locations, then summarize rain, wind, temperature, and calendar-aware weather risks for today. Do not include raw addresses, ZIP codes, coordinates, or calendar event text.

1Connect WeatherWRX as a ChatGPT connector and complete WeatherWRX OAuth.
2Create a scheduled task in ChatGPT if your workspace supports scheduled tasks with connector or tool access.
3Paste the prompt, choose the scheduled send time, and approve WeatherWRX tool access when ChatGPT asks.
4If scheduled tasks cannot use connectors in your workspace, make the task a reminder to open a WeatherWRX briefing or use the provider-neutral cron pattern.

Implementation notes

Use read-only scopes for forecast, alerts, locations, and dashboard context.
Keep write-capable scopes disabled unless the task is supposed to manage WeatherWRX settings.

OpenAI: Connect from ChatGPT

Gemini CLI cron briefing

Terminal-friendly local automation

Best for: Terminal-friendly local automation
Schedule: 0 7 * * 1-5

Starter prompt

Use WeatherWRX MCP tools to list my locations, then write a concise morning weather report with alerts, rain, wind, temperature, and calendar-weather risks. Do not include raw addresses, ZIP codes, coordinates, or calendar event text.

1Add WeatherWRX to the Gemini CLI mcpServers settings with the WeatherWRX MCP endpoint as httpUrl.
2Run Gemini CLI once and complete /mcp auth weatherwrx so Gemini can store the OAuth grant locally.
3Test the prompt interactively and confirm the WeatherWRX tool calls produce the expected briefing.
4Add the command to cron, launchd, GitHub Actions, or another scheduler that runs in the user's local time zone.

Implementation notes

Redirect output to email, Slack, a note file, or the notification channel your scheduler already uses.
Run the command under the same OS user that completed Gemini CLI OAuth.

# Run once after adding WeatherWRX to ~/.gemini/settings.json
gemini
/mcp auth weatherwrx

# Example weekday cron at 7:00 AM local time
0 7 * * 1-5 gemini -p "Use WeatherWRX MCP tools to list my locations, then write a concise morning weather report with alerts, rain, wind, temperature, and calendar-weather risks. Do not include raw addresses, ZIP codes, coordinates, or calendar event text." >> "$HOME/weatherwrx-morning-report.txt"

Gemini CLI: MCP server integration Gemini CLI: Non-interactive usage

Claude API scheduled digest

Team or serverless scheduled reports

Best for: Team or serverless scheduled reports
Schedule: Daily server cron at 7:00 AM

Starter prompt

Use WeatherWRX to prepare a morning operations digest. Prioritize active alerts, locations with formula incidents, wind and precipitation risk, and any calendar-aware conflicts for today. Keep the answer short and omit raw addresses, ZIP codes, coordinates, calendar IDs, and event text.

1Use Claude custom connectors for interactive setup and validation, or manage the WeatherWRX OAuth grant inside your own app for scheduled runs.
2Store refresh tokens and access tokens in your server-side secret store, not in client code or cron text.
3At run time, refresh the WeatherWRX token, pass WeatherWRX as an MCP server to Claude, and ask for the digest prompt.
4Send the final digest through your normal email, Slack, incident-management, or dashboard channel.

Implementation notes

Claude chat connectors are useful for testing the prompt before putting the same workflow behind a scheduler.
Log job status and provider errors only; do not persist raw WeatherWRX tool responses unless your retention policy allows it.

// Pseudocode for a scheduled worker
const weatherwrxToken = await getWeatherwrxMcpAccessToken(userId);

await claude.messages.create({
  mcp_servers: [
    {
      type: "url",
      name: "weatherwrx",
      url: "https://weatherwrx.com/mcp",
      authorization_token: weatherwrxToken
    }
  ],
  messages: [
    {
      role: "user",
      content: "Use WeatherWRX to prepare today's morning weather digest."
    }
  ]
});

Claude: Custom connectors Claude API: MCP connector

Provider-neutral cron job

Custom LLM clients, n8n, Zapier, Workers, or Vercel Cron

Best for: Custom LLM clients, n8n, Zapier, Workers, or Vercel Cron
Schedule: Daily before planning starts

Starter prompt

Create a morning WeatherWRX report for the authorized user's monitored locations. Include active alerts, forecast risks, formula-triggered incidents, and calendar-aware weather conflicts. Keep it concise, actionable, and sanitized.

1Register or configure an MCP-capable client that supports Streamable HTTP and OAuth authorization-code with PKCE.
2Complete WeatherWRX OAuth once, then store the refresh token in the scheduler's encrypted secret store.
3On each run, refresh the token, connect to the MCP endpoint with Authorization: Bearer <token>, and request the briefing tools.
4Ask the selected LLM to summarize the tool results, deliver the message, and record only sanitized job metadata.

Implementation notes

Use one WeatherWRX grant per user so revocation and audit trails remain clear.
Add write scopes only for automations that intentionally change locations, alert formulas, or Calendar Cabinet subscriptions.

1. Refresh the WeatherWRX OAuth access token.
2. Connect the MCP client to https://weatherwrx.com/mcp.
3. Call read-only WeatherWRX tools for locations, alerts, forecast, and briefing context.
4. Ask the LLM to turn the tool results into the user-facing digest.
5. Send the digest and log only sanitized delivery status.

MCP authorization specification

OAuth and scopes

Review permissions before granting access

WeatherWRX scopes describe exactly what a connected assistant can read or draft. Choose read scopes for briefings, forecast questions, and calendar context. Add write scopes only when you want the assistant to help manage WeatherWRX configuration.

weatherwrx.locations.read

weatherwrx.forecast.read

weatherwrx.alerts.read

weatherwrx.manifest.read

weatherwrx.calendar_cabinet.read

weatherwrx.calendar_cabinet.write

weatherwrx.locations.write

weatherwrx.alert_formulas.read

weatherwrx.alert_formulas.write

weatherwrx.infographics.read

weatherwrx.primary_calendar.read

Privacy model

WeatherWRX omits sensitive data by default

MCP responses are scoped to the grant and avoid raw account, location, and calendar identifiers by default. Forecast tools use WeatherWRX cache-backed service logic and may refresh provider weather data server-side when cached data is missing, expired, or stale.

No email addresses

No street addresses or ZIP codes

No raw coordinates

No calendar UUIDs or Google account IDs

No raw query strings

No primary-calendar event descriptions, URLs, or attendees; event titles appear only in Forecast Briefing collision results under the primary-calendar scope

Connection details

Protected resource metadata: https://weatherwrx.com/.well-known/oauth-protected-resource
Authorization server metadata: https://weatherwrx.com/.well-known/oauth-authorization-server
Authorize: https://weatherwrx.com/oauth/authorize
Token: https://weatherwrx.com/oauth/token

Need help connecting a client?

Send the support team the client name, the setup step that failed, and any non-secret error code shown by the MCP host.

Contact Support