Loading ToolBench
Vital signs for your MCP servers: health score, token cost, tool-confusion, migration readiness.
This is a quality-assessment tool for MCP servers, not a production MCP server itself. The single tool 'search_docs' has a minimal schema with only a 'query' parameter (type: string, no description). The tool description is adequate (47 chars) but the parameter lacks a description entirely. The tool is defined in a test file (tests/servers/reference_server.py) and appears to be a reference implementation for testing mcpvitals itself, not a real service. No error handling, no output schema documentation, no pagination, no security considerations. The codebase is a linter/analyzer for MCP servers, not a functional MCP server meant for production use.
Search the documentation for a query and return the top passage.
Parameter 'query' has no description. LLMs cannot infer what kind of query is expected (free text, keywords, structured syntax, etc.).
No output schema documented. Callers do not know what structure to expect from the response (string, object, array, etc.).
Tool is defined in a test file (tests/servers/reference_server.py), not in a production server module. This appears to be a reference implementation for testing the mcpvitals analyzer itself.
No error handling or recovery guidance. If a query returns no results or fails, the LLM has no guidance on what to do next.
No pagination or result limits documented. If documentation is large, returning all results could exhaust context windows.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 47 | 2026-07-28+ | v2 |