SiteProof MCP has 5 tools with basic schemas and descriptions, but significant gaps prevent production use. All tools have descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization), and input schemas are present but minimal. No output schemas documented. Error handling is absent, no recovery guidance, categorization, or actionable error messages. Tool names are verb-noun compliant (scan_*, get_*), but descriptions lack WHEN/WHY context. Parameters lack detailed constraints (e.g., 'depth' enum has no explanation of performance tradeoffs). No security considerations documented despite API key usage. Composition is reasonable (single responsibility per tool), but missing chaining IDs in responses and pagination support.

• Document output schemas for all 5 tools. Example for scan_accessibility: {score: number, grade: string, issues: [{rule: string, severity: string, fix: string}]}. This enables LLM to extract and chain results.
• Expand descriptions to 50-200 chars with WHEN/WHY context. E.g., scan_aeo: 'Scan a website for AI Engine Optimization issues (SEO, schema markup, structured data). Use this after scan_accessibility to optimize for AI agent discoverability.' This disambiguates from scan_accessibility.
• Add error handling with recovery guidance. Wrap apiCall() to catch and categorize errors: 'Invalid URL format. Expected https://example.com' (user-fixable), 'Rate limit exceeded. Retry in 60s' (retryable), 'API key invalid' (fatal, contact admin).
• Enhance parameter descriptions with constraints and rationale. For depth: 'quick (default): regex-based scan, ~2s, free tier. deep: browser-based with axe-core, ~15s, Pro only. Choose quick for fast feedback, deep for comprehensive WCAG 2.1 coverage.'
• Add pagination support. Accept limit (1-100, default 20) and offset/cursor parameters. Return total_issues_found and next_offset in response so LLM can iterate through large result sets without context explosion.