Config-driven MCP server for safe, scoped AI database access
ScopeDB provides three well-structured database tools with clear schemas and descriptions. Tool names follow verb_noun convention (db_describe, db_query, db_mutate). Schemas are comprehensive with proper JSON Schema types, enums, and nested objects. However, parameter descriptions lack depth, most are 1-2 sentences without format constraints, ranges, or actionable guidance. Error handling is not visible in the tool definitions. Output schemas are not documented. The server targets a narrow domain (scoped database access) and executes that well, but lacks the polish expected of production tools.
List available tables and columns. Call without args for overview, or with table name for details.
Update data. Only specific columns are writable. Requires WHERE conditions. Writable tables: (dynamically populated from scope)
Query data with structured parameters. Use db_describe first to understand available tables.
Parameter descriptions lack actionable constraints. 'Table to query' does not specify format, valid values, or dependencies. 'Columns to return' does not explain the 'table.col' syntax or when to use it.
Output schemas are not documented. Handlers (handleDescribe, handleQuery, handleMutate) are exported but their return types are not visible in tool definitions. LLMs cannot plan downstream calls without knowing what fields to expect.
No error handling guidance in tool descriptions. If a query fails (invalid table, permission denied, SQL error), the description does not tell the LLM what to do next or how to recover.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | <=2025-11-25 | v2 |
db_mutate description mentions 'Only specific columns are writable' but does not explain how the LLM discovers which columns are writable. Must call db_describe first? Is there a separate tool?
No pagination or result limit guidance. db_query accepts 'limit' but does not specify min/max bounds or default. Large result sets could exhaust context windows.