Strava MCP server has basic tool definitions with schemas and descriptions, but significant quality gaps limit production readiness. All 4 tools have input schemas and descriptions present, but descriptions are in Japanese (non-standard for LLM optimization), lack actionable context, and are too brief (avg ~40 chars). Parameters lack descriptions in most cases (e.g., 'page', 'per_page' in get_activities have descriptions, but 'athlete_id' in get_athlete_stats lacks context on how to obtain it). No output schemas documented. Error handling is minimal, only basic required-field checks. No tool annotations (readOnlyHint, etc.). Tool composition is reasonable (single responsibility), but missing pagination guidance and result limits. Security: credentials properly injected via environment variables (good), but no audit logging or permission gates.
アスリートのアクティビティ一覧を取得します。ページネーションに対応しています。
特定のアクティビティの詳細情報を取得します。
認証されたアスリートのプロフィール情報を取得します。
アスリートの統計情報(総距離、総時間など)を取得します。
Descriptions are in Japanese and too brief (<50 chars). LLMs trained on English struggle with non-English descriptions and cannot infer when to select tools. Descriptions should be 50-200 chars in English, explaining WHAT, WHEN, and any prerequisites.
Parameter descriptions missing or incomplete. 'page' and 'per_page' in get_activities have descriptions, but 'athlete_id' in get_athlete_stats lacks guidance on how to obtain it. Every parameter needs a description explaining its purpose, format, and constraints.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields without knowing the response structure. Document what fields each tool returns (e.g., get_activities returns array of {id, name, distance, moving_time, ...}).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 47 | 2026-07-28+ | v2 |
No pagination guidance or result limits documented. get_activities accepts page/per_page but lacks guidance on max results, total count, or when to stop paginating. Per pattern:paginated-result, tools returning lists should document limits and offer next_cursor or total_count.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All 4 tools are read-only but lack explicit annotation. Annotations help LLMs reason about safety and retry logic.