A read-only MCP server that performs web searches using DuckDuckGo without opening result URLs. Uses only Windows PowerShell/.NET with no filesystem, shell, browser-control, or credential tools.
Single tool 'search_web' has a well-structured schema with proper type definitions, min/max constraints, and defaults. Description is clear and includes appropriate safety warnings (unverified results, not suitable for critical decisions). Parameter descriptions are present and specific. However, output schema is not documented in the visible code, the tool returns results but the structure of those results is not formally specified. Error handling is present but generic. Tool naming follows verb_noun convention correctly. Overall solid for a single-tool server, but lacks output schema documentation and richer error guidance.
Search the web for information. Results are unverified and may contain inaccurate information. This tool does not open pages and is not suitable for financial, medical, legal, or safety-critical decisions.
Output schema not documented. Tool returns search results but the structure (fields, types, array format) is not formally specified in code or comments. LLMs cannot plan downstream operations without knowing what fields to extract.
Error messages are generic ('Search provider content exceeded the parsing time limit'). No guidance on recovery actions or retry strategy. LLM cannot determine if error is retryable or fatal.
No rate-limit documentation or guidance. Tool enforces MaxSearchesPerMinute=10 and MaxSearchesPerHour=60 server-side, but these limits are not exposed in tool description or error responses. Agent cannot plan around rate limits.
STDIO transport only. Server is not remotely accessible and cannot be used by hosted MCP clients. Limits deployment scenarios significantly.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |