MCP Gateway exposes 32 tools with significant definition gaps. Most tools have basic descriptions (10-50 chars), but parameter documentation is sparse. Many tools lack input schemas entirely or have incomplete type definitions. Error handling is present but generic. Tool naming follows verb_noun convention reasonably well, but composition issues exist: admin_call_function and admin/direct_tool_call appear to duplicate functionality. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. Output schemas are undocumented. This is a typical community server with functional but under-specified tools.

• Add tool annotations to all tools: use readOnlyHint=true for list/get/search tools, destructiveHint=true for delete tools, idempotentHint=true for reload/refresh tools. This enables LLMs to reason about safety without reading descriptions.
• Document output schemas for all tools. Specify which fields are returned (e.g., admin_list_functions returns {tools: [{name, description, parameters}], count}). Include chaining IDs (agent_id, tool_id, server_id) so downstream tools can reference results.
• Consolidate admin_call_function and admin/direct_tool_call into a single canonical tool with clear naming (e.g., admin_invoke_tool). Document the difference between MCP tools, delegation tools, and local functions in the description.
• Complete input schemas for record_research_insight, get_research_insights, create_meeting_from_dsl. Define required/optional parameters, types, and constraints. If these are placeholders, remove them or document their status.
• Expand descriptions for update tools (admin_update_agent, admin_update_protocol) to list which fields are mutable and provide examples. E.g., 'Update agent description, assigned_mcp_servers, or assigned_functions. Pass only the fields to change.'