MCP server for interacting with When2Meet scheduling services
Server has 4 tools with basic schemas and descriptions, but significant gaps in parameter documentation, output schema clarity, and error handling. Tool names follow verb_noun convention (get_, generate_, parse_, mark_) which is good. However, descriptions lack actionable context for LLM selection, parameter descriptions are sparse or missing, and output schemas are not formally documented. The parse-availability-selections tool has incomplete schema visibility in the provided code. Error handling returns generic messages without recovery guidance. No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles.
Creates a structured selection prompt with all available time slots and their timestamps.
Extracts event information from a When2Meet URL including name, dates, and available time slots.
Marks the selected time slots as available on When2Meet using browser automation.
Converts selection codes or direct timestamps into actual When2Meet timestamps. Handles multiple selection formats including day codes, time-of-day patterns, and direct timestamps.
Output schemas not documented. Tools return objects with fields like 'eventDetailsString', 'dayGroups', 'slotLookup', 'selectionPrompt' but no formal schema definition. LLMs cannot plan downstream calls or extract required fields reliably.
Parameter descriptions missing or incomplete. 'eventDetails' param in generate-availability-prompt lacks description of required nested structure. 'promptData' in parse-availability-selections has minimal guidance. LLMs cannot infer what fields are required.
Credentials exposed as parameters. 'mark-when2meet-availability' accepts 'userName' and 'password' as tool parameters. These will be logged in agent traces and prompt history. Must use server-side secret injection.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 47 | 2026-07-28+ | v2 |
No tool annotations. 'mark-when2meet-availability' is destructive (WRITE risk) but lacks destructiveHint annotation. 'get-event-details' and 'parse-availability-selections' are read-only but lack readOnlyHint. Agents cannot distinguish safe from risky tools.
Error handling lacks recovery guidance. Errors return generic messages like 'Error retrieving event details: {message}' without suggesting next steps. No error classification (retryable vs user-fixable vs fatal).