MCP server for Kharkiv Metro Route Planner providing metro routing, schedules, and station information
Server provides 4 tools with schemas and descriptions visible in the source. Naming follows verb_noun conventions (get_route, get_schedule, list_stations, find_station). However, descriptions are minimal (20-60 chars), parameter descriptions lack detail on constraints and formats, output schemas are not documented, and error handling guidance is absent. The server does not implement any error recovery patterns or validate inputs against constraints. Parameters are present and mostly typed, but lack LLM-optimized descriptions. This is a fair-to-poor implementation that requires significant improvement to production quality.
Find station by name (fuzzy search)
Find route between two metro stations
Get train schedule for a station
List all metro stations
Output schemas not documented. LLMs cannot infer what fields get_route, get_schedule, list_stations, and find_station return. This forces agents to guess how to parse responses and risks extraction errors.
Tool descriptions are too brief (20 - 60 chars) and lack context. 'Find route between two metro stations' does not explain WHEN to use this tool vs find_station, what format the output takes, or whether the time parameter affects the returned route. Descriptions should be 50 - 200 chars and answer: What? When? What does it return?
Parameter descriptions lack format constraints. 'Departure time in HH:MM format' is stated in the description, but no regex pattern or explicit format constraint is provided. 'Day type' enum is present, but day_type description does not explain the auto-detect default or how it is determined. 'Language' enum exists but description does not explain which language to prefer when station names are ambiguous.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-21 | D | 52 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 23 | - | v1 |
No error handling or recovery guidance. Tools do not indicate which errors are retryable, what invalid input looks like, or how to self-correct. If station names are mistyped, agents have no guidance to try find_station first.
Pagination not documented. list_stations may return many results; no limit, offset, or total_count parameters are visible. If 3 lines × N stations → large response, context window risk is unmanaged.
get_schedule direction parameter description is vague: 'Direction (terminal station, optional)'. Does this mean pass a station name, or a direction like 'northbound'? Unclear semantics invite invalid input.
No tool-chaining documentation. If find_station resolves a user's station name to a canonical name, does it return the name in the format get_route expects? Are IDs returned? Agents cannot compose tools confidently without knowing what IDs/names are shared.