DNS-AID MCP server demonstrates solid tool design with complete schemas, clear descriptions, and proper risk classification. All 6 tools have descriptions (avg 120 chars), input schemas with typed parameters, and enum constraints where appropriate. Tool names follow verb_noun pattern (publish_agent, discover_agents, verify_agent, list_agents, delete_agent, get_server_info). However, output schemas are not documented in the provided source, parameter descriptions lack format/constraint details (e.g., domain format, port range), and error handling guidance is absent. No evidence of idempotency markers or confirmation patterns for destructive operations (delete_agent). Composition is clean, each tool has single responsibility and clear chaining via domain/name/protocol parameters.

• Document output schemas for all tools. Example: publish_agent returns {success: bool, record_ids: [string], ttl: int}. discover_agents returns {agents: [{name, endpoint, port, capabilities, protocol}], total_count: int}.
• Add format/range constraints to parameter descriptions: 'domain (FQDN, e.g. example.com)', 'port (1-65535, default 443)', 'ttl (60-86400 seconds, default 3600)', 'capabilities (array of strings, e.g. ["read", "write"])'.
• Add dry-run parameter to delete_agent: 'dry_run (bool, default false), if true, validate deletion without executing. Prevents accidental agent removal.'
• Include error guidance in descriptions: 'publish_agent: Returns error if domain is invalid, backend unreachable, or agent already exists. Try verify_agent to check current state.'
• Add pagination to discover_agents and list_agents: 'limit (int, 1-100, default 20), max agents to return. next_cursor (string, optional), for pagination. Response includes total_count and next_cursor if more results available.'

Inferred effective spec: 2026-07-28+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry