An MCP server that provides tools for exploring and querying PostgreSQL databases, including metadata inspection, spatial/GIS operations, and SQL execution. Integrates with Ollama LLM and FastAPI backend.
Server has 15 tools with consistent naming (verb_noun pattern) and all have descriptions. However, descriptions are often terse (10-50 chars), parameter descriptions are sparse or missing, and output schemas are not documented. Input schemas are present but minimal, most tools accept 0-2 params with basic type info. Error handling is present but generic. No tool annotations (readOnlyHint, destructiveHint). Composition is sound: tools are single-purpose and chainable. Security: all tools are READ_ONLY, no secrets exposed. Baseline: 549 production tools average 194 chars for descriptions; this server averages ~80 chars. Parameter descriptions are largely absent, violating the 100% A+ baseline.
Return the number of rows in a table.
Return complete information about a table.
Execute a read-only SQL query for analytical aggregation, ranking, custom joins, or cross-entity questions not fully supported by a dedicated tool. Prefer dedicated tools for direct record retrieval.
Find spatial features near a stored spatial entity. The reference entity is resolved from its actual database table and all distance calculations are performed directly against its stored PostGIS geometry. This supports geometry-to-geometry proximity queries including: - Point -> Point - Point -> LineString - Point -> Polygon - LineString -> LineString - LineString -> Polygon - Polygon -> Polygon Distances are calculated using PostGIS geography casts so returned distances are expressed in metres.
Return all columns of a table.
Return capability level and supported/unsupported features for the database.
Parameter descriptions missing or minimal. Most tools (list_tables, list_schemas, get_geometry_columns, etc.) have parameters with only type info, no description text. Rubric baseline: 100% of A+ tools have parameter descriptions.
Tool descriptions are terse (10 - 50 chars). Rubric baseline: average 194 chars (p10=34, p90=392). Most metadata tools lack context on WHEN to use them vs alternatives (e.g., when to call describe_table vs get_columns). Descriptions should explain prerequisites and guide LLM selection.
Output schemas not documented. No tool specifies what fields are returned or their types. LLMs cannot plan downstream calls or extract data reliably without knowing response structure. Rubric baseline: 100% of A+ tools have documented return types.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 66 | 2025-06-18+ | v2 |
Return a structured catalog of schemas, tables, columns, relationships, spatial tables, indexes, row estimates, and capabilities.
Return all foreign keys of a table.
Return PostGIS geometry columns, geometry types, and SRIDs.
Return indexes for a table.
Return primary key columns of a table.
Return configured semantic mappings separate from physical metadata.
Return every non-system schema present in the PostgreSQL database.
Return every table present in the PostgreSQL database.
Return sample rows from a table.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All tools are READ_ONLY but lack explicit annotation. Current MCP spec (2026-07-28) supports tool annotations for clarity; their absence is a missed opportunity for protocol alignment.
execute_sql description is vague about validation. It says 'read-only' but does not explain how the server enforces this or what happens if a user attempts a write query. Error handling for SQL injection or malicious queries is not documented.