This is a template/example repository for the exray web scraping framework, not a functional MCP server. Tools are documented as TypeScript template files with Chinese descriptions, but lack proper MCP tool registration, JSON Schema input definitions, and structured output schemas. No tool definitions are explicitly registered via MCP protocol. Descriptions exist but are domain-specific (web scraping patterns) rather than LLM-optimized. Input parameters are mentioned in prose but not formalized as JSON Schema with types. Output structures are implied but not documented. This appears to be educational/reference material, not a production MCP implementation.

• Formalize all tool definitions as explicit MCP tool registrations with name, description, inputSchema (JSON Schema), and outputSchema. Use the MCP SDK to register tools in a server initialization function.
• Convert parameter documentation from prose to JSON Schema. For each tool, define inputSchema with properties, types, descriptions, and constraints (enum, minLength, pattern, etc.). Example: 'url' → {type: 'string', description: 'Target URL to fetch (required)', format: 'uri'}.
• Document output schemas for every tool. Define the structure of return values (e.g., detail-page returns {title: string, description: string, canonical: string, image: string, published_at: string | null, word_count: number}). Include field descriptions.
• Rewrite tool descriptions for LLM clarity. Follow the pattern: 'WHAT: [action]. WHEN: [use case]. RETURNS: [key fields]. EXAMPLE: [one-liner].' Keep to 50 - 200 characters. E.g., 'Extract metadata (title, description, image, publish date) from a detail page snapshot. Use after fetching a product or article page.'
• Add error handling guidance to each tool description. Document failure modes and recovery steps. E.g., 'login-session: If login fails, verify credentials and try again. If success_selector is not found, the page may have changed, update the selector.'