Coordinates in, readable forecast out, plus state-level alerts - and the same endpoints answer over plain REST, so a script and an MCP client can use one deployment.
A Spring AI MCP server for weather data with two tools: a readable multi-day forecast for a latitude and longitude, and the active weather alerts for a US state. It runs over HTTP with an SSE endpoint, or over stdio for editor integrations.
- A multi-day forecast in readable prose, for a latitude and longitude
- Active weather alerts for a US state
- The same two calls exposed as ordinary REST endpoints, for callers that are not MCP clients
Java 21 and a current Maven to build the JAR, then Docker Compose to run it - build first, or Compose has nothing to start. The server listens on its configured port with the MCP endpoint at /sse; one environment switch turns it into a stdio server for an editor that wants that instead. Alerts cover US states only.
