A Model Context Protocol server for MySQL database integration, providing tools to execute SQL queries, retrieve database schemas, analyze table structures, and manage database relationships with security controls.
This MySQL MCP server has critical definition quality gaps. Tool descriptions are present but minimal (10-30 chars), falling well below the 50-200 char baseline for LLM-optimized descriptions. Input schemas are visible in code but lack proper JSON Schema structure with type declarations and parameter descriptions. No output schemas are documented. Parameter descriptions are missing entirely, e.g., execute_query accepts 'query' with only 'SQL查询语句' (4 chars in English equivalent), providing no guidance on format, length limits, or injection risks. Error handling is generic ('success': false, 'error': message) with no recovery guidance. Security concerns are unaddressed in tool definitions despite SQL injection risks. Tool composition is reasonable (5 focused tools), but naming could be clearer, 'analyze_data' is vague about what analysis is performed.
分析表数据
执行SQL查询
获取数据库模式信息
获取表之间的关系
获取表结构
Parameter descriptions missing or trivial. 'query' param in execute_query has only '执行SQL查询语句' (4 chars); no guidance on format, length limits, or SQL injection prevention. LLMs cannot infer constraints.
Output schemas not documented. Tools return Dict[str, Any] with no specification of fields, types, or structure. LLMs cannot plan downstream calls or extract required data.
Tool descriptions are minimal (10-30 chars), far below 50-200 char baseline. 'Get table structure' and 'Get schema info' lack context on WHEN to use, WHAT is returned, and dependencies. Insufficient for LLM selection.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 33 | <=2025-11-25 | v2 |
SQL injection risks unaddressed in tool definitions. execute_query accepts raw SQL with only client-side validation (security.validate_query). No mention of parameterized queries, escaping, or safe patterns in tool description.
Error handling is generic. Responses return {'success': false, 'error': message} with no recovery guidance. LLMs cannot determine if error is retryable, user-fixable, or fatal.