ServiceNow MCP has well-structured schemas and clear naming conventions (verb_noun pattern: sn_query, sn_get_record, sn_update_record, sn_create_record, sn_get_widget). All 5 tools have descriptions and input schemas with typed parameters. However, descriptions are minimal (10-50 chars), lacking context on WHEN to use each tool or dependencies between them. Parameter descriptions are sparse, 'table' and 'sys_id' lack guidance on format or discovery. Output schemas are not documented, forcing LLMs to infer response structure. Error handling is absent from visible code. The setup.py skill provides workflow guidance but is not integrated into tool definitions themselves.

• Expand tool descriptions to 100-200 chars. Include WHEN to use each tool: 'Use sn_query to search multiple records with filters; use sn_get_record when you have a specific sys_id.' This disambiguates tool selection.
• Document output schemas for all tools. E.g., sn_query returns {records: [{sys_id, fields...}], total_count, has_more}. This enables LLMs to chain tools correctly.
• Add parameter descriptions with format guidance: 'table (string): ServiceNow table name, e.g. incident, change_request. Call sn_query with empty query to discover available tables.' This reduces discovery calls.
• Add tool annotations: mark sn_update_record and sn_create_record with destructiveHint=true; mark sn_query and sn_get_record with readOnlyHint=true. This guides agent reasoning about side effects.
• Integrate error recovery guidance into tool descriptions: 'If no records found, try broadening the query filter or check table name with sn_query(table=sys_db_object).' This prevents dead-end failures.