Model Context Protocol inspector - a tool for inspecting and testing MCP servers
MCP Inspector exposes 2 tools for introspection and tool invocation. Both tools have descriptions and basic input schemas, but critical gaps limit production readiness. listTools lacks required parameters entirely (only optional _meta). callTool has required parameters (name, arguments) but descriptions are minimal (under 50 chars each). Neither tool documents output schemas, error handling, or recovery guidance. Parameter descriptions do not explain constraints, formats, or valid ranges. No tool annotations (readOnlyHint/destructiveHint) despite callTool being destructive. Schemas present but incomplete, missing type constraints, enums, and detailed field documentation. Error handling is absent; no guidance on what to do if a tool call fails or if a tool name is invalid.
Call a tool on a connected MCP server with specified arguments
List available tools from a connected MCP server
callTool lacks destructiveHint annotation despite being a WRITE tool that invokes arbitrary remote tools. LLMs cannot distinguish safe from unsafe calls without explicit tool annotations.
No output schemas documented for either tool. LLMs cannot plan downstream operations or extract required fields (e.g., tool names from listTools, result from callTool) without knowing response structure.
Parameter descriptions are under 50 characters and lack actionable detail. 'Arguments to pass to the tool' does not explain format, constraints, or how to structure nested objects. 'Optional metadata' does not explain what metadata is supported.
No error handling or recovery guidance. If callTool invokes a non-existent tool or receives invalid arguments, there is no documented error response or guidance on what the LLM should do next.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2025-06-18+ | v2 |
callTool's 'arguments' parameter accepts any object but lacks schema constraints. No enum, pattern, or type validation documented. LLMs will pass arbitrary structures without validation.