Server has 9 tools with complete input schemas and descriptions. Naming follows verb_noun convention (scrape, screenshot, google_search, etc.). Descriptions are detailed (100-250 chars) and explain what each tool does. However, output schemas are not documented, LLMs cannot predict response structure. Parameters lack validation guidance (e.g., no min/max for numeric fields beyond wait). Error handling is minimal; responses are passed through without recovery guidance. No tool annotations (readOnlyHint, destructiveHint) despite all tools being read-only. API key exposure as optional parameter is acceptable but could be clearer about server-side injection.

• Document output schemas for all 9 tools. Specify the structure of responses (e.g., scrape returns {html: string, status: number}, screenshot returns {contentType: string, body: Buffer}). This enables LLMs to plan multi-step workflows.
• Add readOnlyHint: true to all tool definitions in the MCP server registration. This signals to clients that these tools are safe to call without user confirmation.
• Enhance error handling in handleScrape, handleScreenshot, and search handlers. Return structured errors with actionable guidance: 'Invalid country code. Must be ISO 3166-1 alpha-2 (e.g., us, gb, in). Got: xyz.'
• Add min/max constraints to numeric parameters in schema definitions. E.g., results: z.string().regex(/^\d+$/).describe('Number of results per page (1-100).'), or use z.number().min(1).max(100) for quality.
• Expand x_profile and x_post descriptions to 100+ chars. Explain what data is returned and when to use each tool. E.g., 'Fetch a Twitter/X user profile including username, follower count, bio, and verification status. Use this to get user metadata before fetching posts.'