A Model Context Protocol server that provides tools to query and interact with Excel Online data sources via JDBC connections, enabling SQL queries, table discovery, and column introspection.
This server has moderate definition quality with three read-only discovery and query tools. All tools have descriptions and registered schemas, but several key quality gaps reduce confidence. Descriptions are adequate but could be more LLM-optimized. Parameter descriptions exist but lack detail on constraints, formats, and dependencies. Output schemas are not documented, critical for an agent planning downstream calls. Error handling is minimal, with no recovery guidance or structured error responses. The tools follow a sensible verb_noun pattern (get_*, run_*) and are appropriately simple in responsibility, but lack the polish of production-grade tools.
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 are not documented. The tools return CSV-formatted text, but the LLM has no formal schema describing what fields/columns will be present, their types, or how to parse the output. This forces the LLM to infer structure from free-text responses, increasing parsing errors and context waste.
Parameter descriptions lack constraint information. The 'sql' parameter in run_query says 'The SELECT statement to execute' but does not formally specify which SQL clauses are supported, identifier quoting rules, or limits. The description mentions 'Valid clauses: FROM, INNER JOIN, LEFT JOIN, GROUP BY, ORDER BY, LIMIT/OFFSET' inline, this should be a formal constraint or enum, not buried in narrative text.
Error handling is non-existent. The code shows `catch (Exception ex) { throw new RuntimeException("ERROR: " + ex.getMessage()); }`. This returns an unstructured error string to the MCP client with no guidance on recovery, no categorization (retryable vs. fatal), and no actionable next steps. An LLM receiving 'ERROR: Connection refused' cannot self-correct.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 15 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 36 | - | v1 |
Descriptions for optional parameters (catalog, schema in get_tables and get_columns) do not explain when they are needed. Are they always optional? Do they affect behavior in subtle ways? The current text 'The catalog name' is too terse, it should explain 'Optional. Filters results to a specific catalog; leave empty to list all catalogs (if supported).'
No pagination or result limiting is documented. The tools accept no limit or offset parameters, and descriptions do not mention caps on result size. If a data source has thousands of tables or columns, the CSV output could exhaust the context window. Production tools should enforce limits and return pagination cursors.
The run_query description is vague about prerequisites. It says 'Use the `run_query` tool to get a list of available tables', but run_query executes a SELECT, it does not list tables. This appears to be a copy-paste error from get_tables. The description should say: 'Use get_tables and get_columns first to discover available tables and columns, then write a SELECT statement.'
No documentation of tool composition or chaining. The tools are designed as a sequence (get_tables → get_columns → run_query), but descriptions do not guide the LLM on the expected workflow. A preamble in the server description or tool docstrings should explain: 'Call get_tables first to discover available tables, then get_columns to inspect a table, then run_query to fetch data.'