Only1MCP is a proxy aggregator tool with significant definition quality gaps. While tool names are reasonable (list, call, read, subscribe, get, createMessage), descriptions are generic and lack context. Input schemas are minimal or missing entirely for most tools. The `tools/call` tool accepts only a 'name' parameter with no indication of how arguments are passed, creating ambiguity about tool composition. Resource and prompt tools are present but descriptions do not explain when to call them vs. alternatives. The `sampling/createMessage` tool is defined but lacks parameter documentation (input schema is empty {}). Error handling guidance is not visible in any tool description. Output schemas are undocumented, there is no specification of what fields agents should expect from responses. This server appears to be a proxy dispatcher rather than a user-facing tool set, which reduces the need for LLM-optimized descriptions, but the current definitions would not meet production quality standards for downstream agent interaction.

• Add explicit parameter descriptions to all 8 tools. For tools/call, document: (1) the 'name' parameter should be the name of a backend tool (string), (2) a separate 'args' parameter (object) for tool arguments, or (3) a clear mechanism for passing arguments to backend tools.
• Document output schemas for every tool. Example: tools/list should return { tools: [{ name: string, description: string, backend_server: string }], total: number, has_more: boolean }.
• Expand tool descriptions from generic 20-60 char summaries to 100-200 char LLM-optimized descriptions that answer: WHAT does it do? WHEN should the agent call it? WHAT fields are in the response? Example: 'List all tools available across configured MCP backend servers. Call this first to discover what tools are available, then use tools/call to invoke them. Returns array of tools with names, descriptions, and which backend server hosts each one.'
• Split tools/call into multiple composable operations: (1) tools/call_by_name(name: string, args: object) for direct invocation, (2) tools/validate(name: string, args: object) for pre-flight validation (dry-run pattern).
• Remove or re-design sampling/createMessage. Server-initiated Sampling was deprecated in MCP 2025. If the proxy needs to request user input during tool execution, implement it via MRTR (Multi Round-Trip Requests) with result type 'input_required' in the MCP protocol layer, not as a callable tool.
• Add pagination parameters and response structure to resources/list and tools/list. Include: limit (1-100, default 20), offset/cursor, total_count, has_more bool.

Inferred effective spec: <=2025-11-25.

• Relies on Sampling (deprecated) - integrate directly with the LLM provider API