MCP server for searching and fetching documentation from LangChain, LLaMA Index, and OpenAI docs using Serper API
Single tool 'get_docs' has a reasonable description (194 chars, within baseline 34-392 range) and clear naming. However, the input schema lacks formal type declarations and descriptions in the JSON Schema format, parameters are inferred from Python type hints only. No output schema is documented. The tool accepts free-form 'library' strings instead of an enum, inviting hallucinated values. Error handling is minimal (ValueError for unsupported library, generic 'No results found'). No guidance on recovery or retry logic. The tool does one thing (search docs), which is good for composition, but lacks the rigor expected of production tools.
Search the latest docs for a given query and library. Supports langchain, openai, and llama-index. Returns truncated results to work with token limits.
Input schema lacks formal JSON Schema with type and description fields. Parameters 'query' and 'library' are inferred from Python type hints; no explicit schema object visible in tool registration.
'library' parameter accepts free-form strings instead of an enum. Should constrain to ['langchain', 'openai', 'llama-index'] to prevent hallucinated values and self-document valid options.
No output schema documented. LLM cannot predict return structure (string vs object, fields, pagination). Should document: returns string (max 2000 chars) or structured object with 'content', 'source_url', 'library' fields.
Error handling is minimal and non-actionable. ValueError for unsupported library and generic 'No results found' give LLM no recovery guidance. Should return: 'Library not supported. Valid options: langchain, openai, llama-index. Try a different library or refine your query.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | 2026-07-28+ | v2 |
No timeout or rate-limit documentation. Tool calls external APIs (Serper, web fetch) with 30s timeouts but does not document failure modes or retry guidance for LLM.