MCP server for SAP Concur data access via JDBC, providing tools to query tables, retrieve columns, and execute SQL statements
This server provides 3 data-access tools with basic schema definitions and descriptions. However, several critical gaps prevent a higher score: (1) Output schemas are completely undocumented, the tools return CSV text but there is no schema describing the structure LLMs should expect; (2) Parameter descriptions are present but minimal (10-20 chars), lacking format guidance, constraints, or examples; (3) No error handling guidance, tools throw generic RuntimeExceptions with no recovery hints; (4) Tool names follow verb_noun convention (get_tables, get_columns, run_query) which is good, but descriptions are generic and lack context on WHEN to use which tool or how they chain together; (5) The run_query tool description incorrectly references itself ('Use the `run_query` tool to get a list of available tables') creating confusion. Overall, the definitions are functional but lack the precision and LLM-optimization needed for production use.
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.
Output schemas completely undocumented. All three tools return CSV-formatted text, but there is no schema definition describing field names, types, or structure that LLMs should expect. This forces LLMs to infer output structure, increasing parsing errors and wasting tokens on trial-and-error interpretation.
Parameter descriptions are too brief (10-20 chars) and lack actionable detail. For example, 'The table name' does not explain expected format (quoting rules), constraints (max length), or whether it's case-sensitive. Compare to baseline: avg param description is 72 chars.
run_query tool description references itself incorrectly: 'Use the `run_query` tool to get a list of available tables' should reference get_tables instead. This is confusing and misleading.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 40 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 47 | - | v1 |
No error handling guidance. Tools throw generic RuntimeException('ERROR: ' + ex.getMessage()) with no recovery hints. When a table doesn't exist or SQL is malformed, LLMs receive no guidance on what to try next (e.g., 'Call get_tables() first to see available tables').
No input validation or constraint documentation. run_query accepts arbitrary SQL; descriptions do not specify allowed clauses, forbidden keywords (DELETE, INSERT, UPDATE), or SQL dialect limits. This invites LLMs to attempt unsupported operations.
Tool composition clarity is weak. Descriptions mention 'Use the X tool' but do not explain when you would call get_tables vs get_columns vs run_query in sequence, or what field names and IDs chain between them. This forces LLMs to guess the intended workflow.
CSV output format is mentioned but not structured. Tools say 'The output of the tool will be returned in CSV format, with the first line containing column headers', but there is no schema defining what columns are returned or their types. LLMs cannot reliably parse unstructured text output.