Server defines 4 tools with clear verb-noun naming (search_, get_, list_). All tools have descriptions (avg 95 chars, within baseline 34-392). Input schemas present with type declarations. However, critical gaps: (1) Most parameters lack descriptions, 'limit' and 'offset' in search_incidents have minimal guidance; (2) No output schema documentation, callers cannot predict response structure; (3) No enum constraints on parameters like 'limit' ranges (1-200, 1-50) are stated in descriptions but not enforced in schema; (4) Error handling is minimal, graphql() throws generic errors without recovery guidance; (5) No parameter validation rules documented (e.g., date format YYYY-MM-DD is stated but not validated). search_incidents and get_incident are well-named and described, but lack parameter-level detail. list_recent and list_taxonomies are minimal. Overall: functional but below production baseline for agent-facing tools.

• Add output schema documentation for each tool. Example for search_incidents: 'Returns {incidents: [{incident_id: number, title: string, date: string, description: string}], total: number}'.
• Add minimum/maximum constraints to schema for 'limit' and 'offset'. Change inputSchema to include 'minimum': 1, 'maximum': 200 for limit in search_incidents.
• Expand parameter descriptions. For 'query', add: 'Free-text search across incident titles and descriptions. Regex special characters are escaped. Returns up to [limit] results sorted by date (newest first).'
• Improve error messages. Replace 'AIID error: 500 ...' with 'AIID API returned 500. This is a temporary service issue. Retry in 30 seconds or contact support.'
• Add input validation for date strings. In searchIncidents(), validate start_date and end_date match YYYY-MM-DD before sending to GraphQL. Return 'Invalid date format: expected YYYY-MM-DD, got [value]'.