Local MCP server for IMGW-PIB Polish weather, hydro data and warnings — no auth, no cloud middle.
Five tools with clear verb-noun naming (get_*, list_*) and detailed descriptions (150-250 chars each). All tools have input schemas with typed properties and parameter descriptions. However, output schemas are not documented, LLMs cannot predict response structure. Error handling is minimal (no recovery guidance, no categorization). Parameter descriptions lack format constraints (e.g., 'case-insensitive substring' is vague; should specify max length, character set). The hydro tool's computed 'alarm_status' field is well-documented, but no guidance on how to handle API errors or missing data. Schemas are present but incomplete: no required fields declared, no examples, no output type documentation.
Get current hydrological readings from IMGW. Includes water level (stan_wody), flow rate, water temperature, and the alarm / warning threshold levels for the station. Server annotates each with `alarm_status` = normal / warning / alarm computed from current vs thresholds.
Get active IMGW meteorological warnings across Poland. Fields: nazwa_zdarzenia (event type — burze, mróz, upał, opady, wiatr, etc.), stopien (severity 1..3), prawdopodobienstwo (probability %), obowiazuje_od / _do, teren (affected area).
Get current synoptic weather observations from IMGW. With no station: returns all ~60 stations. With station: returns just that one. Data updates ~hourly. Fields: temperature (°C), wind speed (m/s), wind direction (degrees), relative humidity (%), precipitation sum (mm), pressure (hPa).
List hydrological station names, rivers, and provinces from the /hydro feed. Optional query filters. Used to find the right station name for get_hydro.
List synoptic station names available in the /synop feed. Useful for finding the right station name.
Output schemas not documented. LLMs cannot predict response structure (fields, types, nesting). Responses are JSON-dumped as plain text without schema guidance.
No error handling or recovery guidance. API errors (404, timeout, malformed response) return raw JSON without actionable next steps for the LLM.
Parameter constraints are vague. 'case-insensitive substring' lacks max length, character set, or validation rules. LLMs may pass invalid input without clear feedback.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
No pagination or result limits documented for list_* tools. Returning ~900 hydro stations in one response risks context window exhaustion.
get_hydro's 'limit' parameter defaults to 100 but description says 'full set is ~900'. No guidance on how to iterate through all results or what happens when limit < total.