A Model Context Protocol (MCP) server providing comprehensive access to Singapore property prices, HDB data, land use zoning, and neighborhood amenities.
Server has 8 tools with mostly complete schemas and descriptions. Tool names follow verb_noun convention (search_*, analyze_*, export_*, get_*). Descriptions are present and contextual (avg ~150 chars), meeting baseline. However, several parameters lack descriptions or constraints. Output schemas are not documented. Error handling is minimal, tools return empty results or generic errors without recovery guidance. No tool annotations (readOnlyHint/destructiveHint). Missing idempotency guarantees for export tools. Composition is reasonable: search tools chain well, but analyze_results depends on implicit state (last search results), creating hidden coupling.
Analyze the last search results using AI. Provides insights, patterns, and summaries. Best used after a search returns results — works with land use, HDB resale, and amenity data.
Export the last search results to a CSV file. The file must be saved within a directory approved by the client.
Export the last search results to a Markdown file. The file must be saved within a directory approved by the client.
Show data source attributions and licences for all data used by this server. Only call this tool when the user explicitly asks about data sources, credits, or attributions.
Search land use and zoning info near a Singapore address. Resolves the address to coordinates, then finds nearby land parcels with zoning details.
Search land use and zoning info near specific coordinates. Skips geocoding — use when you already have lat/lon.
analyze_results depends on implicit state (last search results) with no explicit input. This creates hidden coupling and makes the tool non-idempotent. LLMs cannot predict what data will be analyzed without seeing the prior search.
export_csv and export_md lack output schema documentation and do not document what happens on file write failure (permission denied, disk full, invalid path). No recovery guidance for LLMs.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). export_csv and export_md are destructive (overwrite files) but not marked. LLMs cannot distinguish safe reads from risky writes.
search_hdb_resale requires exact uppercase town names ('ANG MO KIO', not 'Ang Mo Kio'). Description warns of this but provides no recovery path. Should accept case-insensitive input or return 'Town not found. Valid towns: ANG MO KIO, BEDOK, ...'
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 52 | <=2025-11-25 | v2 |
Search HDB resale flat transactions by town and optional flat type. Data covers 2017 onwards, sorted by most recent. Returns price, floor area, storey, lease, and location details.
Search for nearby amenities (schools, hospitals, clinics, hawker centres, parks, MRT stations, bus stops, supermarkets, pharmacies) around a coordinate in Singapore. Returns results sorted by distance. Requires coordinates — use search_area first to geocode an address if needed.
Output schemas not documented for any tool. LLMs cannot plan downstream calls or extract fields without trial-and-error. search_area, search_hdb_resale, search_nearby_amenities should document returned field names and types.