Server has 8 well-named tools with consistent verb_noun patterns (get_*, search_*). All tools have descriptions (avg ~100 chars, within baseline 34-392). Input schemas are present and typed for all tools. However, descriptions are terse and lack WHEN/WHY guidance. Output schemas are undocumented, responses are free-text strings rather than structured objects. No error handling guidance, no parameter validation hints, no pagination support despite tools returning lists. Missing tool annotations (readOnlyHint, idempotentHint). Composition is sound (single responsibility per tool), but output formatting wastes tokens on human-readable text instead of structured JSON.

• Convert all tool responses from free-text strings to structured JSON. E.g., get_recent_activities should return {activities: [{id: int, name: str, distance_miles: float, moving_time_sec: int, pace_min_per_mile: str, avg_hr: int|null, type: str}], total_count: int, limit: int}.
• Add output schema documentation to every tool. Use JSON Schema to define return type structure. This enables LLMs to plan multi-step workflows and extract fields reliably.
• Enhance descriptions with WHEN/WHY context. E.g., 'Get recent activities (last N runs/rides). Use this for a quick overview; use search_activities for filtering by date/distance/type.' Keep to 50-200 chars.
• Add readOnlyHint: true to all tool schemas to signal these are safe-to-retry idempotent reads. This helps agents distinguish from destructive operations.
• Implement pagination for list-returning tools. Add offset/limit params to get_recent_activities and search_activities. Return {items: [...], total_count: int, offset: int, limit: int} so agents can fetch large result sets incrementally.