Most tools lack proper input schemas and descriptions. Of 22 tools, only 5 have visible Pydantic schemas (list_mcp_servers, search_mcp_tools, add, long_running_task, analyze_sentiment). Many tools show empty input {} with no parameter documentation. Descriptions vary widely: some are detailed (search_mcp_tools at ~280 chars), others are minimal (logging_tool, tool_to_disable). No output schemas documented. Error handling is basic. Tool names follow verb_noun convention well (list_, search_, connect_, disconnect_, get_, add, change_), but composition issues exist: change_tools, change_resources, change_prompts are overly generic and lack clarity on what state changes occur.

• Add Pydantic input schemas to all 17 tools missing them. Even tools with no parameters should declare an empty BaseModel (e.g., class EmptyInput(BaseModel): pass) for consistency.
• Document output schemas for all tools. Specify field names, types, and whether results are paginated. Example: 'Returns {servers: [{name: str, tool_count: int}], total: int}'.
• Expand descriptions for logging_tool, tool_to_disable, purchase_item, start_monitoring, get_current_metrics, get_system_status, ping, get_server_info to 50 - 150 chars. Include WHAT the tool does and WHEN to use it.
• Rename change_tools, change_resources, change_prompts to disable_logging_tool, disable_get_config_resource, disable_summarize_text_prompt (or similar) to clarify what state changes occur.
• Add error handling guidance to all WRITE tools (connect_to_mcp_server, disconnect_from_mcp_server, change_*). Document: 'Returns error if server not found. Try list_mcp_servers() first.' or 'Idempotent, safe to retry.'

Inferred effective spec: 2025-06-18+.

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