MCP server that provides weather tools, resources, and prompts for fetching current weather conditions, forecasts, and location search
This server has a solid foundation with clear tool names, reasonable descriptions, and proper JSON Schema definitions. However, it suffers from several quality gaps that prevent it from reaching production-ready status. All three tools are explicitly defined with input schemas and descriptions, which is a strong baseline. Output schemas are undocumented (not visible in the code), error handling is minimal, and the descriptions, while present, lack depth on WHEN to use each tool and WHAT the output contains. Parameter descriptions are functional but generic. The tools lack hints about their side effects or constraints. For a weather server, this is acceptable but not excellent, typical of a C-grade community implementation.
Get current weather conditions for a specific location
Get weather forecast for a specific location for multiple days
Search for location names matching a query string
Output schemas not documented. LLMs cannot plan downstream calls or extract fields without knowing what the response contains. Current code returns JSON via TextContent but structure is opaque.
Tool descriptions lack WHEN to use hints and do not explain what output structure the LLM should expect. 'Get current weather conditions' is too minimal, should specify return fields (temp, condition, humidity, etc.) and when to call vs forecast.
No error handling guidance. Tool implementations return plain error strings, but do not categorize errors as retryable, user-fixable, or fatal. LLMs receive 'Error: location parameter is required' with no guidance on next steps.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 52 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 48 | - | v1 |
Parameter 'days' in get_weather_forecast has constraints (1-5) correctly declared in schema, but the description 'Number of days to forecast (1-5)' is bare text. Schema constraints are not visible to LLMs, they must parse the natural language description.
Resource read handlers (read_resource) return JSON error strings for invalid URIs, but these are not wrapped in structured error responses. When a resource URI is malformed, the LLM receives a plain JSON object with an 'error' field, no indication of whether to retry or escalate.
search_locations tool lacks outcome guidance. Does it return exact matches, partial matches, or ranked results? How many results? Can pagination be used? Description does not clarify expected output size or structure.