MCP server for PingOne data source connectivity via JDBC, providing tools to query tables, columns, and execute SQL statements
The PingOne MCP server has three well-registered tools with basic schemas and descriptions, but falls short of production quality in multiple critical areas. All tools have descriptions and input schemas (positively), but descriptions lack depth and error handling is minimal. The descriptions are generic rather than LLM-optimized, parameter explanations are sparse, and there is no documentation of output schemas. Tools follow the verb_noun naming pattern correctly but descriptions do not explain WHEN to use each tool versus similar alternatives, nor do they mention prerequisites or dependencies. No error recovery guidance is provided. The server correctly uses Java MCP SDK for registration, but the implementation exposes raw CSV output without structured schemas, forcing LLMs to parse unstructured text. Overall, the tools are functional but underspecified for reliable LLM integration.
Retrieves a list of fields, dimensions, or measures (as columns) for an object, entity or collection (table). Use the `{prefix}_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 `{prefix}_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 not documented. Tools return CSV text but do not formally specify the structure (column names, data types, pagination). LLMs must parse unstructured output, increasing errors and token waste.
Descriptions lack LLM selection context. No explanation of WHEN to use get_tables vs get_columns, whether to call get_tables before get_columns (implied but not stated), or what SQL dialect/limitations apply. Generic descriptions do not guide tool selection.
Error handling is non-existent. Code throws RuntimeException with getMessage(), no categorization (retryable vs user-fixable), no recovery guidance, no actionable error messages. Errors like 'INVALID SQL' will not guide LLM self-correction.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 49 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 51 | 0.8.1+ | v1 |
Parameter descriptions are minimal. 'The catalog name' and 'The schema name' do not explain what happens when omitted (are they optional? defaults?), what format is expected, or when an LLM should provide them. Required 'table' parameter in get_columns has no description at all in schema.
No pagination or result limits. run_query accepts unrestricted SQL SELECT, risking results that blow the context window. No LIMIT enforcement, no offset/limit parameters, no pagination guidance.
SQL dialect constraints buried in description. 'The SQL dialect is mostly based around SQL-92' and allowed clauses are listed as text in the run_query description, not exposed as enums or tool-level validation. LLMs will guess at unsupported syntax.