Three tools with basic descriptions and minimal parameter documentation. All tools have input schemas with types, but descriptions lack LLM-optimization guidance (50-200 char baseline). Parameter descriptions are present but generic. No output schema documentation. Error handling returns raw strings without recovery guidance. Tools are read-only but lack explicit idempotent/readonly annotations. No composition guidance for multi-step workflows.

• Expand tool descriptions to 80-150 chars. Example: 'semantic_query: Search for API endpoints using natural language. Returns endpoint details, auth requirements, and code examples. Use when you need to find how to accomplish a specific API task (e.g., send email, create payment). Requires SEMANTIC_API_KEY.'
• Document output schema for each tool. Specify that responses are JSON objects with fields like 'endpoints', 'auth_type', 'code_examples', 'parameters'. This enables LLMs to extract and chain results.
• Add tool annotations to FastMCP registration: @mcp.tool(readOnlyHint=True, idempotentHint=True) for all three tools to declare safety properties.
• Implement structured error responses with recovery guidance. Instead of 'Error: {e}', return JSON: {'error': 'api_key_missing', 'message': 'SEMANTIC_API_KEY not set', 'recovery': 'Set SEMANTIC_API_KEY environment variable at https://semanticapi.dev'}
• Add composition guidance in descriptions. Example: 'Use semantic_query first for broad discovery, then semantic_discover for deep provider details, then semantic_discover_url to analyze custom API docs.'