Loading ToolBench
Reminders-mcp has clear, action-verb naming (list_*, create_*, complete_*, delete_*) and all 5 tools are explicitly registered with descriptions and input schemas. However, parameter descriptions are minimal (most under 50 chars), output schemas are not documented, and error handling lacks recovery guidance. The server provides basic structure but falls short of production-grade polish. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels.
Mark a reminder as completed.
Create a new reminder in a specific list.
Delete a reminder permanently.
List all Reminder lists available on this Mac.
List reminders in a specific list. Returns incomplete reminders by default.
Output schemas not documented. LLMs cannot predict response structure (fields, types, pagination). Clients must infer from examples or trial-and-error.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Server declares risk levels in metadata but does not expose them via MCP tool annotations, preventing clients from applying safety policies.
Error handling lacks recovery guidance. AppleScript errors are wrapped generically ('AppleScript failed: <msg>') without actionable next steps. LLM cannot distinguish retryable vs fatal failures.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
Parameter descriptions are minimal (10-30 chars). E.g., 'Name of the Reminders list' lacks guidance on list discovery, valid characters, or case sensitivity. Descriptions should be 50-200 chars with examples and constraints.
No pagination support. list_reminders returns all reminders in a single response. Large lists will blow context windows. Should support limit/offset and return total count.