MCP server providing weather information and current time
This server implements two simple tools with basic schemas and descriptions. Both tools are explicitly registered in the ListToolsRequestSchema handler with inline schema definitions. Naming is clear and verb-based (get_weather, get_time). However, there are significant quality gaps: descriptions are below the 194-char baseline for production tools, parameter descriptions are minimal, output schemas are not documented, error handling is minimal (only catches ZodError and MethodNotFound), and the server returns unstructured text responses rather than structured JSON objects. The get_time tool has no parameters at all, which is correct but represents minimal functionality. The get_weather tool accepts a city (required) and units (optional with default 'metric'), but the units parameter description is generic. Most critically, tool outputs are plain text templates rather than structured objects, LLMs cannot reliably extract fields from prose. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) are present. This is typical of early-stage community tools but falls short of production baselines.
Get the current time
Get current weather information for a city
Output schemas not documented. Tools return plain text responses in a 'content' array, but the LLM cannot infer what fields to expect or how to parse results for downstream tool calls. Returning structured JSON objects with typed fields (e.g., {temperature: number, humidity: number, description: string}) would enable reliable field extraction.
Tool descriptions are too brief. 'Get current weather information for a city' (43 chars) and 'Get the current time' (20 chars) are below the 194-char production baseline. Descriptions lack context on when to use the tool, what it returns, and any prerequisites. The get_time description is at the hard cap threshold (<20 chars).
Parameter descriptions are generic or missing. The 'units' parameter description 'Temperature units (metric or imperial)' duplicates the enum constraint; it should explain when to use metric vs imperial and which is the LLM's best default. No description on why 'metric' is the default.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 46 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 45 | - | v1 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Both tools are READ_ONLY and IDEMPOTENT, but these hints are not declared in the tool definitions. This forces LLMs to reason about safety; explicit annotations enable safer agent behavior.
get_weather returns mock data instead of real API calls. The implementation comments 'For demo purposes, we'll return mock data' with random temperature generation. This is acceptable for development, but production deployments require integration with a real weather API (OpenWeatherMap, etc.) and proper error handling for API failures, rate limits, and missing cities.
Error handling is minimal. Only ZodError and MethodNotFound are caught; other errors are unhandled. No actionable error messages guide the LLM on recovery. For example, if the city is invalid or the weather API is unreachable, the agent receives a generic error with no suggestions.
Response objects lack structure. Both tools return a single text field in a 'content' array. Real tool responses should return structured JSON: e.g., {city: string, temperature: number, temperature_unit: 'C'|'F', humidity: number, description: string} for get_weather, and {timestamp: string, timezone?: string} for get_time. This enables the LLM to compose responses and extract values for downstream tools.