An MCP server that matches user questions to FAQ answers using semantic matching
This server exhibits critical quality gaps across all dimensions. The single tool 'find_match' has an extremely generic name that does not follow verb_noun conventions, lacks parameter descriptions, has no documented output schema, and implements a naive word-overlap matching algorithm rather than semantic matching. The implementation hardcodes FAQs (no external data source), lacks error handling guidance, and has no visible MCP server scaffolding, it appears to be a standalone Python class, not an MCP server. No evidence of MCP protocol compliance, transport mechanism, or tool registration visible in the source.
Matches a user question against a set of FAQs and returns the best matching answer
Tool name 'find_match' is vague and does not follow verb_noun convention. 'find_match' does not clearly convey the action or resource; better name: 'search_faq' or 'match_question_to_faq'. The name should make intent obvious without reading the description. [Source: pattern:tool, review:name-clarity]
Input parameter 'user_question' lacks a description. Currently the schema shows only the type and label. [Source: pattern:tool-description]
No documented output schema. The tool returns a dict with keys 'matched', 'question', 'answer', but the MCP tool registration visible in the source does not specify or document these return fields. LLMs cannot plan downstream calls or extract data without knowing the response structure. [Source: pattern:tool, pattern:response-shaper]
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 20 | 2026-07-28+ | v2 |
No error handling or recovery guidance. The tool returns a generic 'I'm sorry, I couldn't find a matching answer.' on no match, which does not guide the LLM on next steps (retry, ask user for clarification, call a different tool).
Algorithm is word-overlap matching, not semantic matching as claimed. The tool name/description promises 'semantic' matching but implements a simple set intersection over lowercased tokens. True semantic matching requires embeddings, cosine similarity, or transformer-based inference. False advertising will disappoint users. [Source: review:incomplete-docstrings]
No visible MCP server scaffolding. The source code shows a plain Python class and example usage, but no evidence of MCP server initialization, tool registration via MCP calls, or connection to an MCP transport (STDIO, HTTP, SSE). This appears to be a standalone Python library, not an MCP server. [Source: protocol spec]
FAQs are hardcoded in the implementation. No mechanism to load FAQs from external source, config file, database, or API. Scaling or updating FAQs requires code changes. Per pattern:tool, tools should accept inputs and delegate to stateless services, not embed data. [Source: pattern:tool, pattern:stateless-operation]
Tool description is only 101 characters and lacks key context. It does not explain when to use this tool vs alternatives, what FAQ sources are available, or what confidence/match threshold is used.