MCP server that provides weather alerts and forecast data using the National Weather Service (NWS) API
This server has 2 tools with basic schemas and descriptions, but falls short of production quality. Both tools have input schemas with reasonable constraints (state length bounds, lat/lon ranges), but descriptions are minimal (under 50 characters each) and lack context about when to use each tool, what they return, or any error handling guidance. Tool names are appropriately verb-based (get-alerts, get-forecast), but no output schemas are documented. The server provides no error recovery guidance, does not indicate which operations are safe to retry, and lacks parameter descriptions explaining what each input controls. This is typical of early-stage educational examples but unsuitable for production agent deployment.
Get weather alerts for a state
Get weather forecast for a location
Tool descriptions are too short and lack context. 'Get weather alerts for a state' (36 chars) does not explain when to use this tool vs get-forecast, what data is returned, or what fields the agent should expect.
No output schema documentation. LLMs cannot plan downstream calls or extract data from results if they don't know what fields to expect. Are alerts returned as a list? Does each alert have a title, timestamp, severity? Unknown.
Parameter descriptions exist for input schemas but are minimal. The 'state' parameter description is 'Two-letter state code (e.g. CA, NY)', acceptable, but examples in descriptions are risky (LLMs reuse them literally). Use enums or patterns instead.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 43 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 41 | - | v1 |
No error handling or recovery guidance. If an invalid state code is passed (e.g. 'XX'), what error does the tool return? How should the LLM recover? Silent failures or opaque HTTP errors provide no guidance.
Tool descriptions do not state whether operations are read-only or idempotent. Both tools are marked risk=READ_ONLY in metadata, but the descriptions do not mention this. Agents need explicit guidance on side effects.
No tool annotations (readOnlyHint, idempotentHint, destructiveHint) in the MCP protocol layer. While the risk metadata is present, the protocol should carry this information in structured form per the current spec.