Single tool 'get_docs' has a basic description but critical gaps in schema completeness, parameter documentation, and error handling. The tool accepts two string parameters (query, library) with minimal constraint documentation. No output schema is documented. Error handling is minimal, only a ValueError for unsupported libraries and a generic 'No results found' string. The description lacks guidance on when to use this tool vs. alternatives, what format results are returned in, or how to handle failures. No pagination, no structured output, no recovery guidance.

• Add enum constraint to 'library' parameter: {"type": "string", "enum": ["langchain", "llama-index", "openai"], "description": "The library to search in. Must be one of: langchain, llama-index, openai."}
• Document output schema: return a structured object with fields like {"results": [{"title": string, "url": string, "snippet": string}], "total_results": number, "library": string}
• Expand description to 150-250 chars explaining: what the tool does, when to use it (e.g., 'Use this to find official documentation for LangChain, OpenAI, or LlamaIndex'), what format results are in, and how to handle no results.
• Add actionable error messages: 'No results found for query "{query}" in {library}. Try a simpler query or check the library name.' Include timeout and API failure guidance.
• Implement pagination: add optional 'max_results' parameter (default 5, max 20) and return structured results with URL, title, and snippet instead of raw text.