A Model Context Protocol server for interacting with a MySQL database
Four tools with basic schemas and descriptions, but significant gaps in parameter documentation, error guidance, and composition. All tools accept only a single 'sql' parameter with minimal constraint documentation. No output schemas documented. Error handling returns text responses without actionable recovery guidance. Write operations (insert, update, delete) lack idempotency guarantees and confirmation patterns. Tool descriptions are adequate (60-90 chars) but do not explain when to use each tool vs. others or what the output structure is. No tool annotations (readOnlyHint, destructiveHint) despite clear risk levels.
Execute a DELETE SQL statement. Requires MYSQL_ALLOW_WRITE_OPS=true.
Execute an INSERT SQL statement. Requires MYSQL_ALLOW_WRITE_OPS=true.
Execute a read-only SQL query against the configured MySQL database.
Execute an UPDATE SQL statement. Requires MYSQL_ALLOW_WRITE_OPS=true.
No output schema documented for any tool. LLMs cannot plan downstream calls or extract structured data. All tools return JSON-stringified rows or error text without field-level documentation.
Write operations (insert, update, delete) lack idempotency guarantees and confirmation/dry-run patterns. Agents retrying on ambiguous failures risk duplicate inserts or unintended data loss. No dry-run mode offered.
Error responses are plain text without actionable recovery guidance. 'MySQL Query Error: ...' or 'Write operations are disabled' do not tell the LLM what to do next or whether to retry.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
Single 'sql' parameter accepts arbitrary SQL with minimal constraint documentation. No description of allowed query types, injection risks, or format expectations. LLMs cannot infer valid SQL syntax from parameter name alone.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Query is read-only; insert/update/delete are destructive. Annotations enable LLM safety reasoning and audit trails.