MCP server for querying SharePoint Excel Services via SQL-like interface
This MCP server presents a moderate-quality SQL/data access interface with clear tool names and basic descriptions, but suffers from significant gaps in parameter documentation, output schema specification, and error handling guidance. The server exposes 3 READ_ONLY tools (get_tables, get_columns, run_query) via a STDIO transport. Tool names are action-oriented and follow verb_noun convention (get_*, run_*), which is good. However, parameter descriptions are minimal (typically 1 - 2 words), output schemas are not formally documented (CSV format is mentioned informally in descriptions but not in structured output definitions), and error handling provides no recovery guidance. The run_query tool description contains a factual error ('Use the `run_query` tool to get a list of available tables', should reference get_tables instead). Parameter definitions show types (string) but lack constraints, format hints, or usage examples appropriate for an SQL-execution interface.
Retrieves a list of fields, dimensions, or measures (as columns) for an object, entity or collection (table). Use the `get_tables` tool to get a list of available tables. The output of the tool will be returned in CSV format, with the first line containing column headers.
Retrieves a list of objects, entities, collections, etc. (as tables) available in the data source. Use the `get_columns` tool to list available columns on a table. Both `catalog` and `schema` are optional parameters. The output of the tool will be returned in CSV format, with the first line containing column headers.
Execute a SQL SELECT statement.
Parameter descriptions are trivially short (1 - 3 words) and lack actionable guidance. 'The catalog name', 'The schema name', 'The table name', 'The SELECT statement' provide minimal context for LLM reasoning. No mention of how these map to database concepts, whether they are case-sensitive, whether they accept wildcards, or what happens if omitted.
Output schemas are not formally documented. Tools return CSV format (mentioned in description text only), but there is no structured specification of which columns are returned, their types, or their order. LLMs cannot reliably extract data or plan subsequent calls without knowing the response structure.
Error handling provides no recovery guidance. Code throws RuntimeException('ERROR: ' + exception.getMessage()), which tells the LLM nothing actionable. No categorization of retryable vs. fatal errors, no suggestions for what to try next (e.g., 'Column not found. Try get_columns() first').
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 44 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 48 | - | v1 |
run_query description contains a factual error: 'Use the `run_query` tool to get a list of available tables, and the `run_query` tool to list table columns.' Both instructions refer to run_query, but should reference get_tables and get_columns respectively. This will confuse LLMs into calling the wrong tool.
No parameter validation guidance or constraints. run_query accepts 'sql' (a string) with no mention of dialect limitations, injection risks, supported clauses, or what happens if malformed SQL is provided. Descriptions mention 'SQL-92' and valid clauses (FROM, JOIN, GROUP BY, LIMIT/OFFSET) but do NOT list them as formal constraints or provide negative examples of invalid usage.
Parameters lack type and format metadata. 'catalog' and 'schema' are marked as strings with no guidance on case sensitivity, null handling, or whether they accept patterns (%, *).
No pagination or result limiting strategy. If a table has thousands of columns or a query returns thousands of rows, the CSV response could exhaust the context window. No limit parameter, no guidance on maximum result size, no warning about performance implications.