Single tool 'search' has a clear action verb and reasonable description (156 chars, within 10-1024 range). Input schema is present with proper JSON Schema structure: 'query' (required, string) and 'search_recency_filter' (optional, enum). However, output schema is NOT documented in the tool definition, the code references ApiResponse interface but does not expose it to the MCP client. Parameter descriptions are adequate but lack detail on expected output structure. No error handling guidance is visible in the tool definition itself.

• Document the output schema in the tool definition. Add a 'returns' field describing: mainContent (string), items (array of {title, snippet, url}), citations (array of strings), media (object). This enables LLMs to extract and chain results.
• Enhance 'search_recency_filter' description: 'Filter results by recency. Options: hour (last 24 hours), day (last 7 days), week (last 30 days), month (last 90 days). Omit for no time filtering. Narrower filters may reduce result count.'
• Add error handling guidance to tool description: 'Returns HTTP 401 if API key is invalid (check AGENTIC_SEARCH_API_KEY env var). Returns HTTP 429 if rate limited (retry after 60s). Returns HTTP 500 for API errors (retryable).'
• Migrate from STDIO to HTTP+SSE or Streamable HTTP transport to enable remote deployment and hosted client compatibility.
• Add per-request logLevel support via _meta field to align with current MCP spec (2026-07-28) for dynamic log control without environment variables.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry