MCP server that provides internet search via SearXNG and webpage reading capabilities, with optional content distillation via Ollama
SearchInlet defines 2 tools with complete input schemas and descriptions. Both tools have verb-noun naming (search, read_page) and clear descriptions (194 and 180 chars respectively). Input parameters include types, descriptions, and constraints (limit 1-50, max_tokens defaults). However, output schemas are not documented, the code returns TextContent but LLMs cannot see the structure of returned fields. Parameter descriptions lack actionable format guidance (e.g., 'URL must be http/https'). Error handling returns generic fmt.Errorf without recovery guidance. No tool annotations (readOnlyHint, idempotentHint) despite both being read-only operations.
Fetch and read the full content of a specific webpage, optimized for LLM context
Search the internet via SearXNG with LLM-optimized output
Output schemas not documented. Code returns TextContent with formatted strings, but LLMs cannot infer the structure of returned fields (title, url, content sections). This forces LLMs to parse unstructured text.
Error handling lacks recovery guidance. Errors like 'searxng search failed' or 'failed to read page' do not tell the LLM what to do next (retry, try different query, check URL format). No error categorization (retryable vs user-fixable).
Missing tool annotations. Both tools are read-only and idempotent, but lack readOnlyHint and idempotentHint annotations. This prevents agents from optimizing retry logic and caching.
Parameter descriptions lack format constraints. 'URL of the page to read' does not specify http/https requirement, URL length limits, or handling of redirects. 'Search query' does not mention length limits or special character handling.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
No pagination support for search results. The 'limit' parameter caps results but does not return a total count or next_cursor. If SearXNG returns 100 results and limit=10, the LLM cannot fetch the next 10 without re-querying.