MCP server for querying Superset databases and executing SQL queries against Superset API
This server has critical definition quality gaps. All three tools lack proper English descriptions (descriptions are in Chinese and very brief), parameters lack type safety and detailed constraints, and no output schemas are documented. The tools are READ_ONLY but lack explicit idempotent hints. The server uses environment variable secrets correctly, but tool definitions themselves are minimal and would confuse LLMs operating in English.
获取所有可用的数据库列表
获取指定数据库的表列表
执行 Superset 数据查询
All tool descriptions are in Chinese (non-English), very short (<20 chars), and provide minimal context. 'Execute Superset data query' alone does not explain WHEN to use this tool vs alternatives, what prerequisites exist, or what it returns.
Output schemas are completely undocumented. Callers cannot infer what fields to expect from query-superset, list-databases, or list-tables responses. This forces LLMs to guess at result structure and risks malformed downstream tool chains.
query-superset parameter 'query' lacks format constraints or examples of acceptable input patterns. The description 'Natural language query, e.g. query latest 10 logs' uses an example value, which LLMs tend to reuse literally rather than generalize.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 38 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 29 | - | v1 |
All optional parameters (databaseId, schema, tableName in query-superset; schema in list-tables) lack guidance on when they should/must be provided and what happens if all are omitted. Undocumented dependencies cause silent misuse.
No error handling guidance in tool descriptions. What happens when a query times out? When a table is not found? When Superset API is unreachable? Descriptions provide no recovery hints.
No tool annotations present. These tools are READ_ONLY and idempotent (safe to retry), but lack explicit readOnlyHint or idempotentHint annotations per MCP spec 2026-07-28. This leaves LLMs uncertain whether tools modify state.
list-databases and list-tables provide no pagination support (no limit, offset, or page parameters; no total_count in output). If Superset has hundreds of databases or tables, results will be truncated or truncate the context window without guidance.