MCP server for Synthetic Web Search API
Single tool with basic schema and descriptions present. Naming is action-oriented (search_web) and follows verb_noun convention well. Description is adequate but generic. Input schema includes query parameter and optional max_text_length with proper type definitions and descriptions. However, output schema is not documented, the tool returns JSON.stringify(results) as text rather than a structured object with typed fields. Error handling exists but lacks recovery guidance. Missing tool annotations (readOnlyHint despite READ_ONLY risk classification). Parameters are minimal and appropriate, but the lack of output schema documentation and structured response design are significant gaps for production use.
Search the web using Synthetic API
Output schema not documented. Tool returns JSON-stringified text rather than structured MCP content with typed fields. LLMs cannot predict result structure for chaining or field extraction.
Tool description (67 chars) is minimal and does not explain WHEN to use this vs other search methods, what structure is returned, or dependencies. Should be 100-200 chars with context.
Error messages are raw API responses without recovery guidance. 'Synthetic API error: 401 ...' tells the agent nothing about how to proceed. Should guide to check API_KEY or retry.
Tool marked as READ_ONLY in risk classification but missing tool annotation. Should declare readOnlyHint: true in schema to inform agents this is safe to call speculatively.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 66 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 33 | - | v1 |
Result truncation at max_text_length uses fragile string slicing. No indication in response whether text was truncated. Should return a 'truncated' boolean per result.