MCP server for accessing Snap4City services API - allows searching and retrieving geolocated services near GPS positions, within geographic areas, by municipality, and full-text search capabilities
Single tool 'get_services' has extremely poor definition quality. Description is excessively long (2,200+ characters, far exceeding the 1024 char baseline for tools), repetitive, and combines 8 distinct use cases into one confusing docstring. Input schema is present but parameter descriptions are inconsistent in depth and quality. Many parameters lack clear format specifications, constraints, and examples. No output schema is documented. The tool violates the single-responsibility principle by bundling multiple search modes (GPS-based, service-based, municipality-based, WKT-based, query-ID-based, full-text) into one function. Parameter naming is inconsistent (selection vs serviceUri; maxDists vs maxResults). No error handling guidance. No pagination support documented despite API returning variable result sets.
Service search near GPS position - It allows to retrieve the set of services that are near a given GPS position. The services can be filtered as belonging to specific categories (e.g. Accommodation, Hotel, Restaurant, etc.), or having specific words in any textual field. It can also be used to find services that have a WKT spatial description that contains a specific GPS position. Service search near a service - It allows to retrieve the set of services that are near a given service identified by its serviceUri. The services can be filtered as belonging to specific categories (e.g. Accommodation, Hotel, Restaurant, etc.), or having specific words in any textual field. It can also be used to find services that have a WKT spatial description that contains a specific GPS position. Service search within a GPS area - It allows to retrieve the set of services that are inside a rectangular area. The services can be filtered as belonging to specific categories (e.g. Accommodation, Hotel, Restaurant, etc.), or having specific words in any textual field. Service search within a WKT described area - It allows to retrieve the set of services that are inside a geographic region described using the Well Known Text (WKT) format. The services can be filtered as belonging to specific categories (e.g. Accommodation, Hotel, Restaurant, etc.), or having specific words in any textual field. Service search within a stored WKT described area - It allows to retrieve the set of services that are inside a geographic region described using the Well Known Text (WKT) format, by referring to the WKT with an identifier provided when the WKT is stored. The services can be filtered as belonging to specific categories (e.g. Accommodation, Hotel, Restaurant, etc.), or having specific words in any textual field. The list of available geometries can be retrieved from the Service Map in the Search Area selection box (with Search Range specific area). New geometries can be provided using the http://www.km4city.org/wkt web service which can store a WKT from a shp file or providing directly the WKT string. Service search by municipality - It allows to retrieve the set of services that are in a specific municipality. The services can be filtered as belonging to specific categories (e.g. Accommodation, Hotel, Restaurant, etc.), or having specific words in any textual field. Service search by query id - It allows to retrieve the set of services associated with a query stored using the Service Map user interface. Full text search - It allows to retrieve the geolocated entities (not only services) that match with a list of keywords. The results can be possibly filtered to be within a specified distance from a GPS position, or within a rectangular area or inside a WKT geolocated area. Service info - It allows to retrieve information about a service using its serviceUri, as an HTML (format query parameter set to html) or a machine-readable JSON document (format query parameter set to json).
Tool description is 2,200+ characters (WAY over 1024 char baseline), repetitive, and conflates 8 distinct search modes instead of breaking them into separate tools. LLMs cannot reason effectively about such bloated, multi-purpose docstrings.
Tool violates single-responsibility principle. 'get_services' performs GPS-near searches, service-near searches, rectangular-area searches, WKT-area searches, municipality searches, query-ID lookups, AND full-text search. Should be split into 4-5 focused tools: search_services_by_location, search_services_by_query_id, search_services_by_municipality, search_services_by_text.
No output schema documented. Callers have no way to know what fields are returned, their types, or whether results are paginated. This forces LLMs to guess the response structure and breaks downstream tool composition.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 33 | 2026-07-28+ | v2 |
Parameter 'selection' description is 600+ chars and describes 6 different input formats (GPS, service URI, rectangular area, WKT string, stored WKT ID, municipality name) with examples for each. This is documentation bloat that should be split across separate tool functions, not crammed into one parameter.
Parameters like 'maxDists', 'maxResults', 'geometry', 'realtime', 'graphUri', etc. have descriptions that are either missing, trivial (<20 chars), or unclear about valid values, format, constraints, or when they apply. LLMs cannot determine which parameters to use for a given search mode.
Parameter 'selection' is required for most use cases but the schema marks it Optional. No validation or error guidance if it is missing or malformed. Similarly, 'search' is described as 'Required' in the text but marked Optional in schema, inconsistency will confuse LLMs.
Parameter 'format' accepts 'json' or 'html' per description but schema has no enum constraint. LLMs may pass invalid values like 'xml' or 'csv'. Should declare format as enum: ['json', 'html'].
Parameter 'lang' described as 'ISO 2 chars language code' but no constraint enforces this. Should either document the exact regex pattern (^[a-z]{2}$) or provide an enum of supported codes (it, en, fr, de, es, etc.).
Parameters 'apikey' and 'accessToken' expose authentication credentials as tool parameters. Per security pattern, credentials must NEVER appear in tool params, they should be injected server-side via environment variables or a secret vault. If present in params, they will appear in agent logs and prompt history.
No pagination support documented. If the API returns hundreds or thousands of services, there's no guidance on how to iterate (page, offset, cursor, limit). The description says maxResults defaults to 100, but doesn't explain what happens if more results exist or how to fetch them.
No error handling guidance. What does the tool return if: selection is malformed? API is unreachable? No results match? Authorization fails? LLMs have no recovery path.